Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces AgenticBBO-Bench, a unified benchmark for evaluating LLM agents in black-box optimization across diverse domains, addressing the lack of comparability in prior studies. Experiments reveal agentic BBO outperforms direct LLM methods and most numerical optimizers, with performance influenced by tool integration, task semantics, and prior knowledge. The benchmark also identifies GPT-6 Astra and DeepSeek-V4.1-Flash as Pareto-optimal models for performance and cost on a frontier challenge.
LLM agents significantly outperform numerical optimizers in four out of five black-box optimization domains, but adding more numerical tools doesn't always help.
Black-box optimization (BBO) arises in many scientific and engineering problems where objective evaluations are expensive and limited. Recent large language model (LLM) agents offer a new way to approach BBO by combining task semantics, computation, optimization tools, and feedback-driven decision making, showing great potential due to the integration with mathematically rigorous tools. However, existing agentic BBO studies use different task domains and system configurations, making their results difficult to compare and the effects of individual design choices hard to isolate. We therefore introduce AgenticBBO-Bench, a cross-domain benchmark for agentic BBO spanning synthetic functions, hyperparameter optimization, database tuning, chip design, and molecular design under a unified finite-budget evaluation protocol. In our experiments, agentic BBO achieves higher family-averaged scores than direct LLM-based methods in all five domains and outperforms the best numerical optimizers in four. We further study three factors shaping agent performance: optimization tools, task information and prior knowledge, and the role of the LLM during search. Our results show that additional numerical tools do not consistently improve performance, task semantics are broadly useful while more specific priors are less reliable, and numerical optimizers can effectively absorb gains from search trajectories established by the agent. Finally, we introduce a five-task frontier challenge within AgenticBBO-Bench and evaluate seven LLMs under the Codex agent harness, where GPT-6 Astra and DeepSeek-V4.1-Flash lie on the Pareto frontier of performance and cost among the evaluated models. Our code is available at https://github.com/lamda-bbo/agentic-bbo.