README.md

metadata

license: apache-2.0
tags:
  - merge
  - mergekit
  - bardsai/jaskier-7b-dpo-v5.6
  - mlabonne/NeuralDaredevil-7B
  - Gille/StrangeMerges_21-7B-slerp
  - CultriX/NeuralTrix-7B-dpo
model-index:
  - name: pikus-pikantny-7B-dare
    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: 72.18
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CorticalStack/pikus-pikantny-7B-dare
          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: 88.56
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CorticalStack/pikus-pikantny-7B-dare
          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: 65
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CorticalStack/pikus-pikantny-7B-dare
          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: 73.29
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CorticalStack/pikus-pikantny-7B-dare
          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: 83.43
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CorticalStack/pikus-pikantny-7B-dare
          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: 70.89
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=CorticalStack/pikus-pikantny-7B-dare
          name: Open LLM Leaderboard

pikus-pikantny-7B-dare

pikus-pikantny-7B-dare is a DARE merge of the following models using mergekit:

• bardsai/jaskier-7b-dpo-v5.6
• mlabonne/NeuralDaredevil-7B
• Gille/StrangeMerges_21-7B-slerp
• CultriX/NeuralTrix-7B-dpo

See the paper Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch for more on the method.

🧩 Configuration

models:
  - model: bardsai/jaskier-7b-dpo-v5.6
  - model: mlabonne/NeuralDaredevil-7B
    parameters:
      density: 0.53
      weight: 0.3
  - model: Gille/StrangeMerges_21-7B-slerp
    parameters:
      density: 0.53
      weight: 0.4
  - model: CultriX/NeuralTrix-7B-dpo
    parameters:
      density: 0.53
      weight: 0.3
merge_method: dare_ties
base_model: bardsai/jaskier-7b-dpo-v5.6
parameters:
  int8_mask: true
dtype: bfloat16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	75.56
AI2 Reasoning Challenge (25-Shot)	72.18
HellaSwag (10-Shot)	88.56
MMLU (5-Shot)	65.00
TruthfulQA (0-shot)	73.29
Winogrande (5-shot)	83.43
GSM8k (5-shot)	70.89