Muse-Autoskill: Self-Evolving Agents via Skill Creation and Memory

The paper introduces Muse-Autoskill, a framework for AI agents to autonomously create, refine, and reuse skills from experience without manual intervention. It uses a skill library and memory system to improve performance on complex tasks by identifying common patterns and reusing successful strategies. Experiments show Muse-Autoskill outperforms baseline methods across various benchmarks.