arxivcs.AI2026-07-15
Experience Memory Graph: One-Shot Error Correction for Agents
Wenjun Wang, Yuchen Fang, Fengrui Liu, Zibo Liang, Kai Zheng
Large Language Model (LLM) agents have shown remarkable capabilities in autonomous decision-making by generating sequential trajectories of states, actions, and observations. However, in complex, long-horizon tasks, these agents frequently suffer from compounding errors and strug…