STaSBO:用 LLM 做策略规划,改进贝叶斯优化的配置选择
STaSBO: LLM-Guided Strategy Planning and Configuration Selection for Bayesian Optimization
把 LLM 当贝叶斯优化的策略大脑,只管选代理和采集函数配置,数值部分照旧跑。在 22 个任务上平均排名压过 11 个基线。做超参调优或实验设计的开发者可以看看这套分工思路。
斯坦福团队(arXiv 2611.19214)提出 STaSBO,一个任务无关的元控制器,用 LLM 为贝叶斯优化(BO)做策略规划并联合选择代理模型-采集函数配置,数值 BO 本身继续负责概率建模和候选点生成。系统分两个阶段:策略规划和策略引导的配置选择,并通过结构化记忆系统把先前的计划、配置选择和候选行为与优化状态关联,控制器能在规划时修订轨迹摘要和行为计划。为应对巨大的联合配置空间,选择阶段先用代理拟合证据做候选短名单,再用前瞻性 BO 动作诊断暴露候选行为,对照行为计划完成最终联合选择。
实验覆盖 12 个合成数学任务和 10 个科学与工程应用任务,STaSBO 在 11 个基线中取得最佳平均排名:两个套件分别为 2.17 和 2.60,而各套件最强基线为 4.08 和 4.10。这说明 LLM 做高层策略、经典数值方法做底层建模的分工式设计,在高成本黑盒优化场景中能稳定优于纯手动或纯自动的配置选择,对超参数调优、材料设计等依赖 BO 的工作流有直接参考价值。
STaSBO: LLM-Guided Strategy Planning and Configuration Selection for Bayesian Optimization
Bayesian optimization (BO) enables sample-efficient search for expensive black-box problems, but its performance depends strongly on surrogate-acquisition configurations whose suitability varies across problems and optimization states. We take a significant step toward LLM-driven strategic management of BO with STaSBO, a task-agnostic meta-controller that plans search behavior and jointly selects surrogate-acquisition configurations, while numerical BO retains responsibility for probabilistic modeling and candidate generation. STaSBO comprises two coordinated stages: strategy planning and strategy-guided configuration selection. A structured memory system connects prior plans, configuration selections, and candidate behaviors to optimization states, allowing the controller to revise its trajectory summary and behavioral plans during planning. To navigate the large joint configuration space, the selection stage shortlists configurations using surrogate fit evidence. Prospective BO-action diagnostics then expose the candidate behaviors, supporting final joint selection against the behavioral plans. Across 12 synthetic mathematical and 10 scientific and engineering application tasks, STaSBO achieves the best average rank against 11 baselines: 2.17 and 2.60, compared with 4.08 and 4.10 for the strongest baseline in each suite, respectively.
本条情报归属 AITOP「论文」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:Bayesian Optimization、LLM、贝叶斯优化、论文,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Youngseok Yoon, Peng Li),经 AITOP 多维评分管线结构化提炼。