EAGLE3-Qwen2.5-VL-7B-Instruct

EAGLE3-Qwen2.5-VL-7B-Instruct (Benchmark Release)

This model repo is part of a multimodal speculative decoding benchmark suite.

Why this repo exists

We maintain a unified benchmark codebase that includes multiple methods (Baseline, EAGLE, EAGLE2, EAGLE3, Lookahead, MSD, ViSpec) so users can run training/evaluation more easily under one setup.

• The methods are aggregated here for user convenience (shared dataset format, scripts, and metrics).
• The original ideas and implementations belong to their respective authors.
• This specific Hugging Face repo hosts the EAGLE3-Qwen2.5-VL-7B-Instruct checkpoint used in our benchmark runs.

Upstream / Base Model

• Base model: Qwen/Qwen2.5-VL-7B-Instruct

What is in this repo

• config.json
• pytorch_model.bin

This checkpoint is intended to be loaded as the EAGLE3 draft/speculative model together with the base model above.

Example usage (benchmark codebase)

bash scripts/Qwen/eval_eagle3_mmspec.sh testmini Cloudriver/EAGLE3-Qwen2.5-VL-7B-Instruct

Method references

• EAGLE: https://arxiv.org/abs/2401.15077
• EAGLE-2: https://arxiv.org/abs/2406.16858
• EAGLE-3: https://arxiv.org/abs/2503.01840
• ViSpec: https://arxiv.org/abs/2509.15235
• Lookahead Decoding: https://lmsys.org/blog/2023-11-21-lookahead-decoding/
• Medusa: https://github.com/FasterDecoding/Medusa

Citation

If you use this checkpoint and benchmark, please cite EAGLE3 and the baseline methods you compare against.

EAGLE / EAGLE2 / EAGLE3

@inproceedings{li2024eagle,
  author = {Yuhui Li and Fangyun Wei and Chao Zhang and Hongyang Zhang},
  title = {{EAGLE}: Speculative Sampling Requires Rethinking Feature Uncertainty},
  booktitle = {International Conference on Machine Learning},
  year = {2024}
}

@inproceedings{li2024eagle2,
  author = {Yuhui Li and Fangyun Wei and Chao Zhang and Hongyang Zhang},
  title = {{EAGLE-2}: Faster Inference of Language Models with Dynamic Draft Trees},
  booktitle = {Empirical Methods in Natural Language Processing},
  year = {2024}
}

@inproceedings{li2025eagle3,
  author = {Yuhui Li and Fangyun Wei and Chao Zhang and Hongyang Zhang},
  title = {{EAGLE-3}: Scaling up Inference Acceleration of Large Language Models via Training-Time Test},
  booktitle = {Annual Conference on Neural Information Processing Systems},
  year = {2025}
}

Notes

• This model card focuses on benchmark usage and attribution.