Jul 2026
SPyCE: Skill-Policy Co-evolution for Multimodal Agents
SPyCE (Skill-Policy Co-evolution), a framework that distills trajectories into a hierarchical skill library and updates it throughout reinforcement learning, is proposed and suggests joint skill-policy optimization as a promising paradigm for building capable multimodal agents.
Ru Zhang, Weijie Qiu
· arXiv.org · 4 citations