arXiv：cs.LG（机器学习，全量分类）· Seongcheol Jeong, Masahiro Suzuki, Yutaka Matsuo·· 14 小时前AI 评分32
SLIM：单纯形格点插值合并，用最少评测优化大模型合并系数
SLIM: Simplex-Lattice Interpolation Merging
AI 导读
研究者提出 Simplex-Lattice Interpolation Merging（SLIM），通过经典混料设计在系数单纯形上构建聚合性能的二次代理模型，仅需评测单个专家模型和等权配对即可确定该代理，之后无需再做目标指标评测即可优化合并系数。
正文
Abstract:Optimizing merging coefficients for large language models can require many costly benchmark evaluations. We propose \textbf{Simplex-Lattice Interpolation Merging (SLIM)}, which constructs a quadratic surrogate of aggregate performance on the coefficient simplex using a classical mixture design. Evaluations of individual experts and equal-weight pairs determine the surrogate with the minimum number of measurements needed to identify a general quadratic on this domain. SLIM then optimizes the surrogate without further target-metric evaluations. Experiments on two model architectures demonstrate accurate prediction of unseen multi-expert mixtures and competitive merge performance under limited evaluation budgets. Matched-budget comparisons show that structured evaluation points improve prediction fidelity over random designs, including those using regularized fitting.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.01037 [cs.LG] |
| (or arXiv:2610.01037v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.01037 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Seongcheol Jeong [view email]
[v1]
Thu, 1 Oct 2026 04:23:54 UTC (1,873 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org