metadata

license: other
license_name: yi-license
license_link: LICENSE
widget:
  - text: 你好! 你叫什么名字!
    output:
      text: 你好，我的名字叫聚言，很高兴见到你。
pipeline_tag: text-generation
model-index:
  - name: OrionStar-Yi-34B-Chat-Llama
    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: 64.93
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OrionStarAI/OrionStar-Yi-34B-Chat-Llama
          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: 84.34
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OrionStarAI/OrionStar-Yi-34B-Chat-Llama
          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: 73.67
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OrionStarAI/OrionStar-Yi-34B-Chat-Llama
          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: 53.35
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OrionStarAI/OrionStar-Yi-34B-Chat-Llama
          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: 78.85
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OrionStarAI/OrionStar-Yi-34B-Chat-Llama
          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: 53.9
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard?query=OrionStarAI/OrionStar-Yi-34B-Chat-Llama
          name: Open LLM Leaderboard

This model is identical to OrionStarAI/OrionStar-Yi-34B with the only difference being that the tensors have been renamed to follow the LLaMA format for automatic evaluation on the HF leaderboard.

Model Introduction

OrionStar-Yi-34B-Chat from OrionStarAI is based on the open-source Yi-34B model, fine-tuned on a high-quality corpus of over 15 million sentences. OrionStar-Yi-34B-Chat aims to provide an excellent interactive experience for users in the large model community.
The Yi series models, open-sourced by the 01-ai team, have shown impressive performance on various benchmarks in Chinese, English, and general domains. OrionStar-Yi-34B-Chat further explores the potential of Yi-34B. Through extensive fine-tuning on a large and high-quality corpus, OrionStar-Yi-34B-Chat performs exceptionally well on evaluation data. We strive to make it an outstanding open-source alternative in the ChatGPT domain!
Our fine-tuned model is completely open for academic research, but please adhere to the agreement and the Yi License.
Model Evaluation Results

We use opencompass to perform 5-shot on the following general domain datasets Testing. The evaluation results of other models are taken from opencompass leaderboard.

	C-Eval	MMLU	CMMLU
GPT-4	69.9	83	71
ChatGPT	52.5	69.1	53.9
Claude-1	52	65.7	-
TigerBot-70B-Chat-V2	57.7	65.9	59.9
WeMix-LLaMA2-70B	55.2	71.3	56
LLaMA-2-70B-Chat	44.3	63.8	43.3
Qwen-14B-Chat	71.7	66.4	70
Baichuan2-13B-Chat	56.7	57	58.4
OrionStar-Yi-34B-Chat	77.71	78.32	73.52

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	68.17
AI2 Reasoning Challenge (25-Shot)	64.93
HellaSwag (10-Shot)	84.34
MMLU (5-Shot)	73.67
TruthfulQA (0-shot)	53.35
Winogrande (5-shot)	78.85
GSM8k (5-shot)	53.90