VCE-Skill: Enhancing Skill Self-Evolution with Version-Change Experience
2026-08-17 • Multiagent Systems
Multiagent SystemsArtificial Intelligence
AI summaryⓘ
The authors study how agents improve their skills for tasks by learning from two sources: past changes made to public versions of skills and new experiences from current tasks. They find these sources complement each other, with public skill changes offering general patterns and current task data providing specific evidence. They introduce VCE-Skill, a method that combines these sources to better evolve skills. Their experiments show VCE-Skill leads to improved performance and better ability to transfer skills across different models.
skill self-evolutionexecution trajectoriesversion historyskill transferpublic skill changestrajectory-derived proposalsknowledge distillationadaptive fusiontask-specific evidence
Authors
Jianming Chen, Xuanbin Ye, Yawen Wang, Junjie Wang, Qing Wang, Fanjiang XU
Abstract
Agents increasingly rely on reusable skills to encode task knowledge, tool-use procedures, and validation rules. Existing skill self-evolution methods primarily revise skills using execution trajectories collected from current tasks, leaving the evolution knowledge accumulated in public skill version histories largely untapped. Our pilot study reveals a clear complementarity between the two sources: public skill changes provide reusable evolution priors, whereas trajectories provide evidence grounded in the current task. Motivated by this, we propose VCE-Skill, which distills noisy and implementation-specific public skill changes into reusable, structured version-change experience and adaptively fuses it with trajectory-derived proposals from the base evolver, thereby exploiting external experience while retaining task-specific evidence. Extensive experiments demonstrate that VCE-Skill improves skill self-evolution, increasing mean scores by 3.20--4.98 points; transfer experiments further show that the resulting skills achieve stronger cross-model transfer performance. Our work highlights public skill version changes as a previously underexplored yet effective source of prior knowledge and advances trajectory-driven skill self-evolution.