ehartford's picture
Adding Evaluation Results (#7)
c44f44a verified
metadata
language:
  - en
license: llama2
datasets:
  - ehartford/WizardLM_evol_instruct_V2_196k_unfiltered_merged_split
model-index:
  - name: WizardLM-1.0-Uncensored-Llama2-13b
    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: 55.72
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ehartford/WizardLM-1.0-Uncensored-Llama2-13b
          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: 80.34
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ehartford/WizardLM-1.0-Uncensored-Llama2-13b
          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: 55.4
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ehartford/WizardLM-1.0-Uncensored-Llama2-13b
          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: 51.44
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ehartford/WizardLM-1.0-Uncensored-Llama2-13b
          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: 74.66
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ehartford/WizardLM-1.0-Uncensored-Llama2-13b
          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: 13.27
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=ehartford/WizardLM-1.0-Uncensored-Llama2-13b
          name: Open LLM Leaderboard

This is a retraining of https://huggingface.co/WizardLM/WizardLM-13B-V1.0 with a filtered dataset, intended to reduce refusals, avoidance, and bias.

Note that LLaMA itself has inherent ethical beliefs, so there's no such thing as a "truly uncensored" model. But this model will be more compliant than WizardLM/WizardLM-13B-V1.0.

Shout out to the open source AI/ML community, and everyone who helped me out.

Note: An uncensored model has no guardrails. You are responsible for anything you do with the model, just as you are responsible for anything you do with any dangerous object such as a knife, gun, lighter, or car. Publishing anything this model generates is the same as publishing it yourself. You are responsible for the content you publish, and you cannot blame the model any more than you can blame the knife, gun, lighter, or car for what you do with it.

Like WizardLM/WizardLM-13B-V1.0, this model is trained with Vicuna-1.1 style prompts.

You are a helpful AI assistant.

USER: <prompt>
ASSISTANT:

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 49.31
ARC (25-shot) 55.72
HellaSwag (10-shot) 80.34
MMLU (5-shot) 55.4
TruthfulQA (0-shot) 51.44
Winogrande (5-shot) 74.66
GSM8K (5-shot) 13.27
DROP (3-shot) 14.35

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 55.14
AI2 Reasoning Challenge (25-Shot) 55.72
HellaSwag (10-Shot) 80.34
MMLU (5-Shot) 55.40
TruthfulQA (0-shot) 51.44
Winogrande (5-shot) 74.66
GSM8k (5-shot) 13.27