README.md

metadata

language:
  - en
license: apache-2.0
model-index:
  - name: mhm-7b-v1.3-DPO-1
    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: 49.57
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-7b-v1.3-DPO-1
          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: 68.1
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-7b-v1.3-DPO-1
          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: 45.76
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-7b-v1.3-DPO-1
          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: 45.88
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-7b-v1.3-DPO-1
          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: 62.04
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-7b-v1.3-DPO-1
          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: 15.24
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=h2m/mhm-7b-v1.3-DPO-1
          name: Open LLM Leaderboard

A DPO fine tuned mhm-7b-v1.3 on Intel/orca_dpo_pairs

Based upon mistral. Created using dare_ties and models from openllm leaderboard. Over 3 merges involving 7 different models, this was the result.

Just an experiment.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	47.77
AI2 Reasoning Challenge (25-Shot)	49.57
HellaSwag (10-Shot)	68.10
MMLU (5-Shot)	45.76
TruthfulQA (0-shot)	45.88
Winogrande (5-shot)	62.04
GSM8k (5-shot)	15.24