Reflexion: Language Agents with Verbal Reinforcement Learning Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials
[NeurIPS 2023] Reflexion: Language Agents with Verbal . . . - GitHub This repo holds the code, demos, and log files for Reflexion: Language Agents with Verbal Reinforcement Learning by Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, Shunyu Yao