Datasets:

evanellis

/

manifold

like 0

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]

{ "Clerk": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": { "profile_fields": [ "Age and demographic details", "Professional background", "Relevant expertise/experience", "Geographic location", "Key perspective they bring" ], "prompt_template": "Generate {num_jurors} diverse jury profiles for a case about: {root_question}\n\nRequirements:\n{requirements}\n\nFor each juror, provide:\n{profile_fields}\n\nFormat as JSON with structure:\n{{\n \"Juror1\": {{\n \"name\": \"Juror1\",\n \"profile\": \"Detailed profile in a few sentences in the second person...\"\n }},\n ...\n}}", "requirements": [ "Ensure demographic diversity (age, ethnicity, geography, profession)", "Include relevant expertise for the topic", "Vary socioeconomic backgrounds", "Mix urban/rural perspectives", "Include different educational levels" ] }, "name": "Clerk", "system_message": "You are a court clerk responsible for jury selection. You ensure diverse representation while selecting jurors with relevant expertise for the case topic. Reply in JSON only, with no accompanying text. Use the second person to describe each juror." }, "Defense": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Defense", "system_message": "You argue for lower probability forecasts based on the evidence." }, "HeadAgent": { "capabilities": { "llm": true, "web_retrieval": true }, "jury_selection": null, "name": "HeadAgent", "system_message": "You are the head agent responsible for gathering information and coordinating other agents." }, "Judge": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Judge", "system_message": "You are a fair and impartial judge overseeing the trial." }, "Prosecutor": { "capabilities": { "llm": true, "web_retrieval": false }, "jury_selection": null, "name": "Prosecutor", "system_message": "You argue for higher probability forecasts based on the evidence." }, "UserAgent": { "capabilities": { "llm": false, "web_retrieval": false }, "jury_selection": null, "name": "UserAgent", "system_message": "You are a user agent that initiates tasks and receives responses." } }

[ { "action_type": "generate_jury", "agent": null, "criteria": "relevance and objectivity", "initiator": "Clerk", "max_turns": 1, "message": null, "num_jurors": 11, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "UserAgent", "silent": false, "step": "0", "store": "" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Given the forecasting question '{root_question}', generate a search query to gather relevant news articles and information. Focus on current evidence and developments related to the question. Only output the search query and NOTHING ELSE.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "1", "store": "search_query" }, { "action_type": "web_search", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 20, "num_top_articles": 6, "query": "{search_query}", "receiver": null, "silent": false, "step": "2", "store": "articles" }, { "action_type": "process_articles", "agent": "HeadAgent", "criteria": "relevance and objectivity", "initiator": null, "max_turns": 1, "message": null, "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": null, "silent": false, "step": "3", "store": "top_articles" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on these articles, provide a concise summary of the key evidence regarding {root_question}. Focus on factual information and current developments, pay attention to the date of the article.\n\nArticles:\n{top_articles}", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "HeadAgent", "silent": false, "step": "4", "store": "evidence_summary" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "We will be evaluating {root_question}. Here is the evidence summary: {evidence_summary}. Please establish the trial rules and procedures.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "5", "store": "trial_rules" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be high. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Prosecutor", "silent": false, "step": "6", "store": "prosecution_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the evidence summary: {evidence_summary}, present your opening argument for why the probability should be low. Include specific evidence and reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Defense", "silent": false, "step": "7", "store": "defense_opening" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Consider the following: Evidence Summary: {evidence_summary}\nProsecution Opening: {prosecution_opening}\nDefense Opening: {defense_opening}\n\nBased on this information, what probability would you assign to the question: {root_question}? Provide your probability estimate (0-100%) and brief reasoning.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Jurors", "silent": false, "step": "8", "store": "jury_forecasts" }, { "action_type": "message", "agent": null, "criteria": "relevance and objectivity", "initiator": "UserAgent", "max_turns": 1, "message": "Based on the jury forecasts: {jury_forecasts}, please provide a final summary and the median probability forecast.", "num_jurors": 3, "num_results": 10, "num_top_articles": 6, "query": null, "receiver": "Judge", "silent": false, "step": "9", "store": "final_verdict" } ]