AutoRefine: From Trajectories to Reusable Expertise for Continual LLM Agent Refinement

Abstract

Large language model agents often fail to accumulate knowledge from experience, treating each task as an independent challenge. Recent methods extract experience as flattened textual knowledge, which cannot capture procedural logic of complex subtasks. They also lack maintenance mechanisms, causing repository degradation as experience accumulates. We introduce AutoRefine, a framework that extracts and maintains dual-form Experience Patterns from agent execution histories. For procedural subtasks, we extract specialized subagents with independent reasoning and memory. For static knowledge, we extract skill patterns as guidelines or code snippets. A continuous maintenance mechanism scores, prunes, and merges patterns to prevent repository degradation. Evaluated on ALFWorld, ScienceWorld, and TravelPlanner, AutoRefine achieves 98.4%, 70.4%, and 27.1% respectively, with 20-73% step reductions. On TravelPlanner, automatic extraction exceeds manually designed systems (27.1% vs 12.1%), demonstrating its ability to capture procedural coordination.

Repository Summary

AutoRefine converts agent trajectories into reusable expertise, separating procedural know-how from static knowledge and continually refining, pruning, and merging these reusable assets to support long-term agent improvement.

Bibliographic Data

Title

AutoRefine: From Trajectories to Reusable Expertise for Continual LLM Agent Refinement

Authors

Libin Qiu, Zhirong Gao, Junfu Chen, Yuhang Ye, Weizhi Huang, Xiaobo Xue, Wenkai Qiu, Shuo Tang

Publication date

2026/01/30

Identifier

arXiv:2601.22758

DOI

10.48550/arXiv.2601.22758

PDF size

1.6 MB