arXiv cs.AI· Xiaotang Feng, Philip Torr, Bruno Andreis·· 14 小时前AI 评分34
Online Surrogate Repair：将高保真反馈与搜索长度解耦的闭环发现算法
Online Surrogate Repair: Decoupling High-Fidelity Feedback from Search Length in Closed-Loop Discovery
AI 导读
研究者提出闭环算法 Online Surrogate Repair（OSR），用稀疏高保真评估在较长的智能体搜索过程中持续更新代理模型，把高保真评估频率与搜索时长解耦。
正文 · 原文
Abstract:Closed-loop AI scientists can generate candidate designs at low marginal computational cost, whereas reliable feedback may require wet-lab synthesis, characterization, or high-fidelity computation. Addressing this imbalance through custom laboratory automation remains infrastructure-intensive and costly, while replacing new experiments with a fixed surrogate leaves persistent model errors that can be amplified by optimization. We propose \emph{online surrogate repair} (OSR), a closed-loop algorithm that uses sparse high-fidelity evaluations to update the surrogate throughout a longer agent search conducted primarily with inexpensive surrogate feedback. An acquisition rule selects which designs from the agent's accumulated proposals receive high-fidelity evaluation, and the resulting labels update the surrogate used in subsequent episodes. Across controlled synthetic environments, we demonstrate that improving global surrogate fit does not necessarily reduce maximum regret, whereas Q90-UCB and expected improvement (EI) substantially reduce regret by directing evaluations toward regions that determine the optimizer's decisions. \osradded{Across LLM and Chemeleon workflows, Online EI reduces the amount of oracle queries required by 84.3--90.3\% for matching discovery counts and by 77.9--81.6\% for matching the quality of the best material found.} Online surrogate repair introduces a third feedback regime between fixed-surrogate operation and high-fidelity feedback after every episode, separating the frequency of high-fidelity evaluation from the duration of the agent's search.
| Subjects: | Machine Learning (cs.LG); Materials Science (cond-mat.mtrl-sci); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.07655 [cs.LG] |
| (or arXiv:2609.07655v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2609.07655 arXiv-issued DOI via DataCite |
Submission history
From: Xiaotang Feng Mr. [view email]
[v1]
Mon, 7 Sep 2026 15:42:45 UTC (1,903 KB)
[v2]
Mon, 28 Sep 2026 21:45:21 UTC (1,907 KB)
[v3]
Thu, 29 Oct 2026 13:58:13 UTC (1,904 KB)
来源：arXiv cs.AI · arxiv.org