一作ACL 2026Findings· Findings of ACL
Agentic Episodic Control
Xidong Yang · Wenhao Li · Junjie Sheng · Yun Hua · Haosheng Chen · Chuyun Shen · Xiangfeng Wang
摘要
Reinforcement learning (RL) remains fundamentally limited by poor data efficiency and weak generalization. Prior episodic RL methods attempt to alleviate this via external memory modules, yet they suffer from two key limitations: a representation bottleneck caused by shallow encoders, and a retri...