AEL: Agent Evolving Learning for Open-Ended Environments

ID: 2604.21725

Authors: Wujiang Xu, Jiaojiao Han, Minghao Guo, Kai Mei, Xi Zhu, Han Zhang, Dimitris N. Metaxas

Focus: Converting episodic experience into persistent behavioral improvement.

Key Insight: Solves the "statelessness" of current agents by introducing a mechanism to convert past episodes into refined future policies, allowing agents to evolve over hundreds of sequential tasks.

RSI Relevance: Bridges the gap between short-term memory and long-term policy evolution in persistent agent systems.

View on ArXiv