Preprint
Aug 2026
Learning What to Remember and What to Internalize in LLM Self-Evolution via Adaptive Memory-Parameter Coordination
COVE is presented, a unified agent self-evolution framework that combines harness-based and parameter-based learning through task-aware routing, stage-aware scheduling, and knowledge optimization, and shows that COVE outperforms single-channel evolution strategies.
T. Ji, Zhenya Huang, Jiayu Liu et al.
· 0 citations