metadata

library_name: transformers
tags:
  - mergekit
  - merge
base_model:
  - bunnycore/Phi-4-Model-Stock
  - bunnycore/Phi-4-rp-v1-lora
model-index:
  - name: Phi-4-Stock-RP
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 63.99
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bunnycore/Phi-4-Stock-RP
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 55.21
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bunnycore/Phi-4-Stock-RP
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 32.25
            name: exact match
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bunnycore/Phi-4-Stock-RP
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 14.43
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bunnycore/Phi-4-Stock-RP
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 18.53
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bunnycore/Phi-4-Stock-RP
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 47.96
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=bunnycore/Phi-4-Stock-RP
          name: Open LLM Leaderboard
license: mit

Phi-4-Stock-RP is a phi4 based language model designed for reasoning and role-play scenarios. It leverages the capabilities of several pre-existing high-quality models, integrating them into a cohesive system that excels in reasoning, creative, narrative, and interactive text generation.

Training Data:

Sources: Merged from various pre-trained models, focusing on those with strong performance in text generation and understanding. Enhanced with a specialized LoRA trained on role-play dialogues, scenarios, and character interactions. Model Capabilities:
Role-Playing: Capable of maintaining coherent characters, plots, and dialogues over extended interactions. Creative Writing: Assists in crafting stories, dialogues, and character development with a focus on immersion and narrative coherence. General Language Understanding: Inherits general text comprehension and generation from the base models, making it versatile for various language tasks beyond RP.

<|im_start|>system<|im_sep|> {system_message}<|im_end|> <|im_start|>user<|im_sep|> {prompt}<|im_end|> <|im_start|>assistant<|im_sep|>

Merge Method

This model was merged using the passthrough merge method using bunnycore/Phi-4-Model-Stock + bunnycore/Phi-4-rp-v1-lora as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:



base_model: bunnycore/Phi-4-Model-Stock+bunnycore/Phi-4-rp-v1-lora
dtype: bfloat16
merge_method: passthrough
models:
  - model: bunnycore/Phi-4-Model-Stock+bunnycore/Phi-4-rp-v1-lora
tokenizer_source: unsloth/phi-4

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	38.73
IFEval (0-Shot)	63.99
BBH (3-Shot)	55.21
MATH Lvl 5 (4-Shot)	32.25
GPQA (0-shot)	14.43
MuSR (0-shot)	18.53
MMLU-PRO (5-shot)	47.96