--- license: other license_name: yi-license license_link: LICENSE model-index: - name: yi-34b-200k-rawrr-dpo-2 results: - task: type: text-generation name: Text Generation dataset: name: AI2 Reasoning Challenge (25-Shot) type: ai2_arc config: ARC-Challenge split: test args: num_few_shot: 25 metrics: - type: acc_norm value: 64.68 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=adamo1139/yi-34b-200k-rawrr-dpo-2 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: HellaSwag (10-Shot) type: hellaswag split: validation args: num_few_shot: 10 metrics: - type: acc_norm value: 84.74 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=adamo1139/yi-34b-200k-rawrr-dpo-2 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: MMLU (5-Shot) type: cais/mmlu config: all split: test args: num_few_shot: 5 metrics: - type: acc value: 75.96 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=adamo1139/yi-34b-200k-rawrr-dpo-2 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: TruthfulQA (0-shot) type: truthful_qa config: multiple_choice split: validation args: num_few_shot: 0 metrics: - type: mc2 value: 46.15 source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=adamo1139/yi-34b-200k-rawrr-dpo-2 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: Winogrande (5-shot) type: winogrande config: winogrande_xl split: validation args: num_few_shot: 5 metrics: - type: acc value: 83.19 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=adamo1139/yi-34b-200k-rawrr-dpo-2 name: Open LLM Leaderboard - task: type: text-generation name: Text Generation dataset: name: GSM8k (5-shot) type: gsm8k config: main split: test args: num_few_shot: 5 metrics: - type: acc value: 61.79 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=adamo1139/yi-34b-200k-rawrr-dpo-2 name: Open LLM Leaderboard --- Anti-refusal anti-instruct capabilities of this model are much stronger than yi-34b-200k-rawrr-dpo-1. This model is Yi-34B-200K fine-tuned using DPO on rawrr_v1 dataset using QLoRA at ctx 500, lora_r 16 and lora_alpha 16. I then applied the adapter to base model. This model is akin to raw LLaMa 65B, it's not meant to follow instructions but instead should be useful as base for further fine-tuning. Rawrr_v1 dataset made it so that this model issue less refusals, especially for benign topics, and is moreso completion focused rather than instruct focused. Base yi-34B-200k suffers from contamination on instruct and refusal datasets, i am attempting to fix that by training base models with DPO on rawrr dataset, making them more raw. You should be able to achieve good 0ctx uncensoredness and quite good lack of gptslop if you finetune this model for instruct. # [Open LLM Leaderboard Evaluation Results](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) Detailed results can be found [here](https://huggingface.co/datasets/open-llm-leaderboard/details_adamo1139__yi-34b-200k-rawrr-dpo-2) | Metric |Value| |---------------------------------|----:| |Avg. |69.42| |AI2 Reasoning Challenge (25-Shot)|64.68| |HellaSwag (10-Shot) |84.74| |MMLU (5-Shot) |75.96| |TruthfulQA (0-shot) |46.15| |Winogrande (5-shot) |83.19| |GSM8k (5-shot) |61.79|