TensorX
返回文献探索

Paper · arXiv 2503.04222

FuseChat-3.0: Preference Optimization Meets Heterogeneous Model Fusion

Ziyi Yang, Fanqi Wan, Longguang Zhong, Canbin Huang, Guosheng Liang, Xiaojun Quan

15 upvotesMarch 6, 2025arXiv 预印本
AI 摘要

FuseChat-3.0, a suite of large language models, integrates diverse source models into smaller targets using supervised fine-tuning and Direct Preference Optimization, achieving significant performance improvements across various tasks.

large language modelsheterogeneous source LLMscompact target LLMsGemma-2-27B-itMistral-Large-Instruct-2407Qwen-2.5-72B-InstructLlama-3.1-70B-Instructsupervised fine-tuningDirect Preference Optimizationinstruction followinggeneral knowledgemathematicscodingAlpacaEval-2Arena-Hard

Abstract

We introduce FuseChat-3.0, a suite of large language models (LLMs) developed by integrating the strengths of heterogeneous source LLMs into more compact target LLMs. Our source models include the powerful Gemma-2-27B-it, Mistral-Large-Instruct-2407, Qwen-2.5-72B-Instruct, and Llama-3.1-70B-Instruct. For target models, we focus on three widely-used smaller variants-Llama-3.1-8B-Instruct, Gemma-2-9B-it, and Qwen-2.5-7B-Instruct-along with two ultra-compact options, Llama-3.2-3B-Instruct and Llama-3.2-1B-Instruct. To leverage the diverse capabilities of these source models, we develop a specialized data construction protocol tailored to various tasks and domains. The FuseChat-3.0 training pipeline consists of two key stages: (1) supervised fine-tuning (SFT) to align the target and source model distributions, and (2) Direct Preference Optimization (DPO) to apply preferences from multiple source LLMs to fine-tune the target model. The resulting FuseChat-3.0 models exhibit significant performance gains across tasks such as instruction following, general knowledge, mathematics, and coding. As illustrated in Figure 1, using Llama-3.1-8B-Instruct as the target model, our fusion approach achieves an average improvement of 6.8 points across 14 benchmarks. Moreover, it demonstrates remarkable gains of 37.1 points and 30.1 points on the instruction-following benchmarks AlpacaEval-2 and Arena-Hard, respectively. Our code, models, and datasets are available at https://github.com/SLIT-AI/FuseChat-3.0.

北京市昌平区探索星信息技术及软件开发工作室

京ICP备2026059466号