🤖 AI Summary
This work addresses a critical limitation in existing autonomous scientific discovery systems, which often rely on experimental memory or heuristic summarization and lack explicit, uncertainty-aware modeling of belief over hypothesis quality. To overcome this, the authors propose BayesEvolve, a novel framework that integrates Bayesian inference with large language models to construct an updatable predictive belief state that actively guides experimental design. Central to this approach is a belief-guided selection mechanism incorporating annealed uncertainty-aware rewards, which substantially improves sample efficiency under a fixed evaluation budget. Empirical results demonstrate that the learned belief state effectively predicts the quality of candidate hypotheses and enables efficient late-stage focused exploration, thereby accelerating the discovery process.
📝 Abstract
Autonomous scientific discovery systems increasingly use large language models (LLMs) to propose new hypotheses, but many such systems condition primarily on experimental memory: archives of high-scoring candidates or heuristic summaries of recent trials. We argue that discovery agents should instead maintain explicit, uncertainty-aware beliefs about hypothesis quality. We introduce BayesEvolve, a belief-guided discovery framework that converts experimental evidence into a predictive belief state and uses this belief to guide future experimentation. As a controlled testbed for belief-guided discovery, we evaluate BayesEvolve on shifted BBOB-style black-box optimization tasks, leaving program and laboratory discovery domains to future work. BayesEvolve improves sample efficiency over memory- and archive-guided LLM baselines under a fixed evaluation budget. We further show that the belief state is predictive on held-out candidate pools, that controlled decision-rule ablations favor belief-guided selection with an annealed uncertainty bonus, and that BayesEvolve exhibits productive late-stage concentration rather than unfocused exploration.