AgentSwift: Efficient LLM Agent Design via Value-guided Hierarchical Search

Yu Li Lehui Li Zhihao Wu Qingmin Liao Jianye Hao Kun Shao, Fengli Xu Yong Li · 2025

Read AgentSwift: Efficient LLM Agent Design via Value-guided Hierarchical Search by Yu Li Lehui Li Zhihao Wu Qingmin Liao Jianye Hao Kun Shao, Fengli Xu Yong Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-06017.txt.

Description

Yu Li and co-authors automate the design of language-model agents with a guided search in a 2025 paper. Hand-built agents combine workflows with memory, planning, and tool use, but automated design has mostly shuffled workflows alone, and testing each candidate agent can cost tens of dollars. Their framework searches a hierarchy that covers both workflow and plug-in components together.

To afford the search, they train a value model on cheap evaluations gathered with a coverage-plus-sampling strategy. A tree search steered by that value model and by uncertainty then explores the space. Across seven benchmarks the discovered agents improved on prior automated designs. The authors frame the lesson broadly: judge candidates cheaply and search hierarchically, and automated design becomes practical.

Catalog record

CREATORYu Li Lehui Li Zhihao Wu Qingmin Liao Jianye Hao Kun Shao, Fengli Xu Yong Li
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-06017

Browse the full library · About the Tower of Babel