Summary
Testing many design ideas in fields like chemistry or materials science can be expensive and slow because each test requires real experiments or detailed computations. The authors created a new approach called online surrogate repair (OSR) that mostly uses fast, approximate tests to explore many options but occasionally checks some promising ideas with accurate tests to update its approximation. This method helps balance speed and accuracy, leading to better decisions while needing fewer costly verifications. Their experiments show OSR can find good designs using much less expensive feedback than previous methods that either always checked accurately or never updated the approximations.
closed-loop discoverysurrogate modelhigh-fidelity evaluationoptimizationacquisition functionexpected improvementmulti-armed banditsmachine learningagent-based searchregret minimization
Authors
Xiaotang Feng, Philip Torr, Bruno Andreis
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. On MADE, controls receiving high-fidelity feedback after every episode require $6.36$--$7.23\times$ more oracle queries to match Online EI under two LLM orchestrators and $10.27\times$ more under the non-LLM Chemeleon+MLIP workflow. Online surrogate repair introduces a novel 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.