UltraQwen-7B / README.md
leaderboard-pr-bot's picture
Adding Evaluation Results
9c8c82f verified
|
raw
history blame
4.07 kB
metadata
language:
  - en
license: other
datasets:
  - HuggingFaceH4/ultrachat_200k
base_model: Qwen/Qwen-7B
model-index:
  - name: UltraQwen-7B
    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: 51.71
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Locutusque/UltraQwen-7B
          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: 77.93
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Locutusque/UltraQwen-7B
          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: 59.16
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Locutusque/UltraQwen-7B
          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: 48.2
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Locutusque/UltraQwen-7B
          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: 73.95
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Locutusque/UltraQwen-7B
          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: 44.05
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=Locutusque/UltraQwen-7B
          name: Open LLM Leaderboard

Model description

The model was trained on about 100,000 examples of the HuggingFaceH4/ultrachat_200k dataset, with plans to release more checkpoints later on.

This model has not been aligned with DPO. In the future, different repositories will be released that contain versions of this model aligned with DPO, using various datasets.

Evaluation

Upon personal testing, the model demonstrates excellent performance in mathematics, history, trivia, and coding tasks. This model can be found on the Open LLM Leaderboard.

Recommended inference parameters

temperature=0.2, top_p=0.14, top_k=12, repetition_penalty=1.1

License

Please make sure to read the Qwen licensing agreement before using this model.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 59.17
AI2 Reasoning Challenge (25-Shot) 51.71
HellaSwag (10-Shot) 77.93
MMLU (5-Shot) 59.16
TruthfulQA (0-shot) 48.20
Winogrande (5-shot) 73.95
GSM8k (5-shot) 44.05