Dream-COP：在经验世界中引导大语言模型发现组合优化算法
Dream-COP: Guiding Algorithm Discovery for Combinatorial Optimization Problems in an Experience World
AI 导读
研究者提出 Dream-COP，利用大语言模型通过迭代程序生成进行组合优化问题的算法发现，核心是将历史搜索轨迹及结果构成“经验世界”，由 policy agent 在其中通过 dreaming 学习可执行策略，引导 discovery agent 生成和改进 COP 程序。
正文 · 原文
Abstract:Large language models (LLMs) enable algorithm discovery for combinatorial optimization problems (COPs) through iterative program generation and evaluation. A central challenge is how to organize search to discover better algorithms within a limited budget. Our key insight is that past search trajectories and their recorded outcomes form an experience world, which can be used to facilitate the COP solving through replay. We introduce Dream-COP, in which a \emph{policy agent} learns executable policies through \emph{dreaming} in this experience world to guide a \emph{discovery agent} in generating and refining COP programs. Specifically, policy replay provides feedback on the quality of the programs revealed and the search steps used, to guide policy selection more effectively. Across five COPs, \textsc{Dream} discovers better algorithms than fixed exploration in both heuristic component design and complete solver generation. Particularly, it matches the optimal values on KP and reduces the gap on OBPP from 2.112\% to 1.199\% while using 42.5\% fewer search requests. Moreover, the learned exploration policies transfer effectively across different COPs with the assitance of the experience world.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.36618 [cs.LG] |
| (or arXiv:2610.36618v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36618 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Kuan Xu [view email]
[v1]
Tue, 27 Oct 2026 04:22:22 UTC (484 KB)
来源：arXiv cs.AI · arxiv.org