README.md

metadata

language:
  - en
license: gemma
datasets:
  - openbmb/UltraFeedback
pipeline_tag: text-generation
model-index:
  - name: Gemma-2-9B-It-SPPO-Iter2
    results:
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: IFEval (0-Shot)
          type: HuggingFaceH4/ifeval
          args:
            num_few_shot: 0
        metrics:
          - type: inst_level_strict_acc and prompt_level_strict_acc
            value: 31
            name: strict accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: BBH (3-Shot)
          type: BBH
          args:
            num_few_shot: 3
        metrics:
          - type: acc_norm
            value: 42.17
            name: normalized accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MATH Lvl 5 (4-Shot)
          type: hendrycks/competition_math
          args:
            num_few_shot: 4
        metrics:
          - type: exact_match
            value: 0
            name: exact match
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: GPQA (0-shot)
          type: Idavidrein/gpqa
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 11.3
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MuSR (0-shot)
          type: TAUR-Lab/MuSR
          args:
            num_few_shot: 0
        metrics:
          - type: acc_norm
            value: 10.94
            name: acc_norm
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2
          name: Open LLM Leaderboard
      - task:
          type: text-generation
          name: Text Generation
        dataset:
          name: MMLU-PRO (5-shot)
          type: TIGER-Lab/MMLU-Pro
          config: main
          split: test
          args:
            num_few_shot: 5
        metrics:
          - type: acc
            value: 31.89
            name: accuracy
        source:
          url: >-
            https://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard?query=UCLA-AGI/Gemma-2-9B-It-SPPO-Iter2
          name: Open LLM Leaderboard

Self-Play Preference Optimization for Language Model Alignment (https://arxiv.org/abs/2405.00675)

Gemma-2-9B-It-SPPO-Iter2

This model was developed using Self-Play Preference Optimization at iteration 2, based on the google/gemma-2-9b-it architecture as starting point. We utilized the prompt sets from the openbmb/UltraFeedback dataset, splited to 3 parts for 3 iterations by snorkelai/Snorkel-Mistral-PairRM-DPO-Dataset. All responses used are synthetic.

Terms of Use: Terms

Links to Other Models

• Gemma-2-9B-It-SPPO-Iter1
• Gemma-2-9B-It-SPPO-Iter2
• Gemma-2-9B-It-SPPO-Iter3

Model Description

• Model type: A 8B parameter GPT-like model fine-tuned on synthetic datasets.
• Language(s) (NLP): Primarily English
• License: Apache-2.0
• Finetuned from model: google/gemma-2-9b-it

AlpacaEval Leaderboard Evaluation Results

Model	LC. Win Rate	Win Rate	Avg. Length
Llama-3-8B-SPPO Iter1	48.70	40.76	1669
Llama-3-8B-SPPO Iter2	50.93	44.64	1759
Llama-3-8B-SPPO Iter3	53.27	47.74	1803

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-07
• eta: 1000
• per_device_train_batch_size: 8
• gradient_accumulation_steps: 1
• seed: 42
• distributed_type: deepspeed_zero3
• num_devices: 8
• optimizer: RMSProp
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_train_epochs: 1.0

Citation

@misc{wu2024self,
      title={Self-Play Preference Optimization for Language Model Alignment}, 
      author={Wu, Yue and Sun, Zhiqing and Yuan, Huizhuo and Ji, Kaixuan and Yang, Yiming and Gu, Quanquan},
      year={2024},
      eprint={2405.00675},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	21.22
IFEval (0-Shot)	31.00
BBH (3-Shot)	42.17
MATH Lvl 5 (4-Shot)	0.00
GPQA (0-shot)	11.30
MuSR (0-shot)	10.94
MMLU-PRO (5-shot)	31.89