Reasoning BO is introduced, a novel framework leveraging reasoning models to guide BO sampling while incorporating multi‐agent systems and knowledge graphs for online knowledge accumulation, and it is demonstrated that smaller LLMs, after post‐training, can achieve performance comparable to larger counterparts.
Abstract
Many real‐world scientific and industrial applications require the optimization of expensive black‐box functions. Bayesian optimization (BO) provides an effective framework but often struggles with local optima and lacks interpretability. This paper introduces reasoning BO, a novel framework leveraging reasoning models to guide BO sampling while incorporating multi‐agent systems and knowledge graphs for online knowledge accumulation. We evaluate our approach across 10 diverse tasks, including synthetic functions and complex real‐world chemical optimizations. Reasoning BO progressively refines sampling strategies through real‐time insights and hypothesis evolution, identifying high‐performing regions effectively. In the direct arylation task, our method significantly outperformed traditional BO, increasing yield from 25.20% to 60.07%. Furthermore, we demonstrate that smaller LLMs, after post‐training, can achieve performance comparable to larger counterparts. This framework establishes an intelligent cost‐effective optimization system for scientific discovery, combining LLM reasoning with structured knowledge management.
This work introduces agentic Bayesian optimization: a paradigm in which an LLM agent is the central decision maker in the BO loop while a Bayesian backend provides the uncertainty-aware optimization substrate, and demonstrates the practical value of agentic BO in dynamic settings.
Paul Brunzema, Louis C. Tiao, Nhat Le et al.· 1 citation
A surrogate-recommendation framework is introduced that predicts the most suitable BO surrogate from inexpensive dataset characteristics and establishes FruBO as a reproducible, compute-aware baseline for Bayesian Optimization and provides practical guidance for surrogate selection under limited computational and experimental budgets.
P. Krokidas, C. Rekatsinas, Vassilis Sioros et al.· arXiv.org· 1 citation
Traditional automated feature engineering (AFE) minimizes human intervention but often neglects semantic relationships among original features, resulting in redundant or uninterpretable transformations. While recent methods leverage the domain knowledge and reasoning capabilities of Large Language Models (LLMs), they typically constrain the search space via predefined mathematical operators or suffer from poor exploration-exploitation balance due to exclusive reliance on validation feedback. To address these limitations, we propose MORE-FE, an evolutionary AFE framework that integrates multi-operator exploration with Reinforcement Learning with Verifiable Rewards (RLVR). MORE-FE treats feature transformation programs as hypotheses and evolves them using evolutionary operators that explore semantically coherent and logically structured features through various prompting strategies. Moreover, it employs RLVR to align LLM reasoning with structured evolutionary exploration using a composite reward that balances quality and diversity. Extensive experiments on multiple classification and regression datasets demonstrate that MORE-FE consistently outperforms state-of-the-art AFE methods, highlighting the effectiveness of balancing quality and diversity in feature engineering.
Chang-Yu Chao, Bryan Andersen, Xiao-Xi Tan et al.· Proceedings of the 32nd ACM...· 0 citations
The nature of test-time exploration in RLVR-trained LLMs is investigated by employing controlled maze-solving experiments and extracting a tree structure from mathematical reasoning traces (BODHI-Trees) based on semantic equivalence to delineate between entropy arising from stylistic variations and genuine inferential branching.
Soumadeep Saha, Krish Sharma, Akshay Chaturvedi et al.· 1 citation
Analysis shows that BRANCH's advantage arises not only from exploring multiple reasoning paths, but also from recovering from truncation: its gains strongly correlate with the baseline rate of empty, budget-exhausted outputs, weakening the hypothesis that different problems require routing among test-time reasoning operators.
Sheng Zhang, Xiao-Min Wu, Xiyang Wu et al.· 0 citations
PoTRE (Poly-Topological Reasoning Ensembles), a heterogeneous framework that decouples inference into four agents that achieves improved reasoning performance using similar or fewer inference tokens compared to heavily scaled homogeneous baselines is introduced.
