metadata

language:
  - en
license: llama2
datasets:
  - umd-zhou-lab/claude2_alpaca
model-index:
  - name: claude2-alpaca-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: 61.18
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=umd-zhou-lab/claude2-alpaca-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: 84.21
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=umd-zhou-lab/claude2-alpaca-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: 55.93
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=umd-zhou-lab/claude2-alpaca-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: 45.02
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=umd-zhou-lab/claude2-alpaca-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: 76.8
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=umd-zhou-lab/claude2-alpaca-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: 28.28
            name: accuracy
        source:
          url: >-
            https://huggingface.co./spaces/HuggingFaceH4/open_llm_leaderboard?query=umd-zhou-lab/claude2-alpaca-13B
          name: Open LLM Leaderboard

Model Card for umd-zhou-lab/claude2-alpaca-13B

This model is trained by fine-tuning llama-2 with claude2 alpaca data.

Model Details

Model Description

Developed by: UMD Tianyi Zhou Lab
Model type: An auto-regressive language model based on the transformer architecture
License: Llama 2 Community License Agreement
Finetuned from model: meta-llama/Llama-2-13b

Model Sources

GitHub: Claude2-Alpaca
Data: claude2_alpaca

Uses

The primary use of this model is research on large language models and chatbots. The primary intended users of the model are researchers and hobbyists in natural language processing, machine learning, and artificial intelligence.

Training

We use the prompt from Stanford Alpaca

Hyperparameter	Global Batch Size	Learning rate	Epochs	Max length	Weight decay
Model (13B)	128	1e-5	5	2048	0

Performance

Compared to the llama2-chat, our models can have better average performance.

	Average	ARC	HellaSwag	MMLU	TruthfulQA	Alpaca_Eval	Avg Length
Llama-2-7b-chat	56.335	52.9	78.55	48.32	45.57	71.37	1479
Llama-2-13b-chat	59.935	59.04	81.94	54.64	44.12	81.09	1513

claude_alpaca-7b	57.78	56.66	81.17	46.58	46.71	71.23	1066
claude_alpaca-13b	61.29	61.18	84.08	55.74	44.18	78.93	1127

Citation

Please consider citing our paper if you think our codes, data, or models are useful. Thank you!

@misc{claude2-alpaca,
  author = {Lichang Chen and Khalid Saifullah and Ming Li and Tianyi Zhou and Heng Huang},
  title = {Claude2-Alpaca: Instruction tuning datasets distilled from claude},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/Lichang-Chen/claude2-alpaca}},
}

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	58.57
AI2 Reasoning Challenge (25-Shot)	61.18
HellaSwag (10-Shot)	84.21
MMLU (5-Shot)	55.93
TruthfulQA (0-shot)	45.02
Winogrande (5-shot)	76.80
GSM8k (5-shot)	28.28