| 📑 FuseLLM Paper @ICLR2024 | 📑 FuseChat Tech Report | 📑 WRPO Tech Report | 🌐 Website |
| 🤗 HuggingFace Repo | 🐱 GitHub Repo |
- Dec 12, 2024: 🔥 We release FuseChat-3.0 and Blog Post. FuseChat-3.0 contains a series of models crafted to enhance performance by integrating the strengths of multiple source LLMs into more compact target LLMs. To achieve this fusion, we utilized four powerful source LLMs: Gemma-2-27b-It, Mistral-Large-Instruct-2407, Qwen-2.5-72B-Instruct, and Llama-3.1-70B-Instruct. For the target LLMs, we employed three widely-used smaller models—Llama-3.1-8B-Instruct, Gemma-2-9B-It, and Qwen-2.5-7B-Instruct—along with two even more compact models—Llama-3.2-3B-Instruct and Llama-3.2-1B-Instruct. . The implicit model fusion process involves a two-stage training pipeline comprising Supervised Fine-Tuning (SFT) to mitigate distribution discrepancies between target and source LLMs, and Direct Preference Optimization (DPO) for learning preferences from multiple source LLMs. The resulting FuseChat-3.0 models demonstrated substantial improvements in tasks related to general conversation, instruction following, mathematics, and coding. Notably, when Llama-3.1-8B-Instruct served as the target LLM, our fusion approach achieved an average improvement of 6.8 points across 14 benchmarks. Moreover, it showed significant improvements of 37.1 and 30.1 points on instruction-following test sets AlpacaEval-2 and Arena-Hard respectively.
-
Aug 16, 2024: 🔥🔥🔥🔥 We update the FuseChat tech report and release FuseChat-7B-v2.0, which is the fusion of six prominent chat LLMs with diverse architectures and scales, namely OpenChat-3.5-7B, Starling-LM-7B-alpha, NH2-Solar-10.7B, InternLM2-Chat-20B, Mixtral-8x7B-Instruct, and Qwen1.5-Chat-72B. FuseChat-7B-v2.0 achieves an average performance of 7.38 on MT-Bench (GPT-4-0125-Preview as judge LLM), which is comparable to Mixtral-8x7B-Instruct and approaches GPT-3.5-Turbo-1106.
-
Mar 13, 2024: 🔥🔥🔥 We release a HuggingFace Space for FuseChat-7B, try it now!
-
Feb 26, 2024: 🔥🔥 We release FuseChat-7B-VaRM, which is the fusion of three prominent chat LLMs with diverse architectures and scales, namely NH2-Mixtral-8x7B, NH2-Solar-10.7B, and OpenChat-3.5-7B. FuseChat-7B-VaRM achieves an average performance of 8.22 on MT-Bench, outperforming various powerful chat LLMs like Starling-7B, Yi-34B-Chat, and Tulu-2-DPO-70B, even surpassing GPT-3.5 (March), Claude-2.1, and approaching Mixtral-8x7B-Instruct.
-
Feb 25, 2024: 🔥 We release FuseChat-Mixture, which is a comprehensive training dataset covers different styles and capabilities, featuring both human-written and model-generated, and spanning general instruction-following and specific skills.
- Jan 22, 2024: 🔥 We release FuseLLM-7B, which is the fusion of three open-source foundation LLMs with distinct architectures, including Llama-2-7B, OpenLLaMA-7B, and MPT-7B.
Please cite the following paper if you reference our model, code, data, or paper related to FuseLLM.
@inproceedings{wan2024knowledge,
title={Knowledge Fusion of Large Language Models},
author={Fanqi Wan and Xinting Huang and Deng Cai and Xiaojun Quan and Wei Bi and Shuming Shi},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/pdf?id=jiDsk12qcz}
}
Please cite the following paper if you reference our model, code, data, or paper related to FuseChat.
@article{wan2024fusechat,
title={FuseChat: Knowledge Fusion of Chat Models},
author={Fanqi Wan and Longguang Zhong and Ziyi Yang and Ruijun Chen and Xiaojun Quan},
journal={arXiv preprint arXiv:2408.07990},
year={2024}
}
Please cite the following paper if you reference our model, code, data, or paper related to WRPO.
@article{yang2024wrpo,
title={Weighted-Reward Preference Optimization for Implicit Model Fusion},
author={Ziyi Yang and Fanqi Wan and Longguang Zhong and Tianyuan Shi and Xiaojun Quan},
journal={arXiv preprint arXiv:2412.03187},
year={2024}
}