作者位次 #5/6ACL 2026Poster· Natural Language Processing
EVOTOOL : Self-Evolving Tool-Use Policy Optimization in LLM Agents via Blame-Aware Mutation and Diversity-Aware Selection
Shuo Yang · Soyeon Caren Han · Xueqi Ma · Yan Li · Mohammad Reza Ghasemi Madani · Eduard Hovy
摘要
LLM-based agents depend on effective tool-use policies to solve complex tasks, yet optimizing these policies remains challenging dueto delayed supervision and the difficulty ofcredit assignment in long-horizon trajectories.Existing optimization approaches tend to beeither monolithic, which are pr...