arXiv cs.CL· Shuo Cai, Yanggan Gu, Yuhang Liu, Su Lu, Yuanyi Wang, Yang Yu, Zhen Li, Zhitian Hou, Guanghao Zhu, Ming Li, Hongxia Yang·· 13 小时前AI 评分34
FusionEval：面向大语言模型的模型融合基准测试
FusionEval: Benchmarking Model Fusion for Large Language Models
AI 导读
研究者提出 FusionEval 基准，用 Qwen3 和 Llama 基座独立训练的数学、代码、科学、指令遵循与工具使用专家池，统一评测有无额外训练的融合方法。该套件在八个公开基准上报告归一化平均与最差性能、运行时间和峰值 GPU 显存。实验显示免训练方法保留推理能力但在两种模型规模下都显著削弱工具使用，MOPD 能保留工具使用且能力更均衡，但需额外数据准备、推理与训练。
正文 · 原文
Authors:Shuo Cai, Yanggan Gu, Yuhang Liu, Su Lu, Yuanyi Wang, Yang Yu, Zhen Li, Zhitian Hou, Guanghao Zhu, Ming Li, Hongxia Yang
Abstract:Model fusion aims to integrate the capabilities of multi-expert models into a single model. Existing evaluations offer limited coverage of reasoning with long chains of thought and tool use, while strong average performance can mask substantial losses in individual expert capabilities. We introduce FusionEval, a benchmark that provides controlled expert pools, integrates fusion methods with and without additional training, and offers a unified evaluation suite. The pools contain domain experts independently trained from Qwen3 and Llama base models across math, code, science, instruction following, and tool use. The suite evaluates capability retention and computational cost on eight public benchmarks under comparable data access rules, reporting normalized average and worst performance, runtime, and peak GPU memory usage. Extensive experiments show that training-free methods preserve reasoning performance but substantially degrade tool use at both model scales. In contrast, MOPD preserves tool use and achieves a more balanced capability profile, but requires additional data preparation, inference, and training. Jointly retaining reasoning, instruction following, and tool use at low computational cost remains a challenge. We hope FusionEval supports future research on efficient model fusion that jointly preserves expert capabilities. Our code is available at this https URL.
| Comments: | 20 pages, 4 figures |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.30162 [cs.CL] |
| (or arXiv:2610.30162v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.30162 arXiv-issued DOI via DataCite |
Submission history
From: Shuo Cai [view email]
[v1]
Wed, 30 Sep 2026 18:10:13 UTC (234 KB)
来源：arXiv cs.CL · arxiv.org