SkillEvoReg: Regularizing Agent Skill Evolution Against Overfitting
Language-model agents increasingly improve by converting execution experience into reusable external skills. Yet repeated skill updates form a learning process of their own: locally useful edits can accumulate into redundant or task-specific instructions, while new updates can disrupt behavior that previously worked. W...
Guan-Yu Nie, Fang-Zhou Zhu, Shi-Xiong Kai et al.
· 0 citations