VCE-Skill - Enhancing Skill Self-Evolution with Version-Change Experience

Sep 20, 2026

Uploaded by Margaret

Agents increasingly rely on reusable skills to encode task knowledge, tool-use procedures, and validation rules. Existing skill self-evolution approaches revise skills from execution trajectories collected on current tasks, leaving public skill version histories underused. This work proposes VCE-Skill, which distills noisy, implementation-specific public version changes into structured version-change experience and adaptively fuses it with trajectory-derived proposals. Experiments show mean score gains of 3.20–4.98 points and improved cross-model transfer.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages23

Reading Time58 min

AccessFree to download

Introduction

Motivation: trajectory limits and public version histories

Proposed method: VCE-Skill with distillation and adaptive fusion

Evaluation results (summarized)

VCE-Skill: Enhancing Skill Self-Evolution with Version-Change Experience Jianming Chen 1,2,3,4 , Xuanbin Ye5 , Yawen Wang 1,2,3,4∗, Junjie Wang 1,2,3,4∗, Yuanzhe Hu 1,2,3,4 , Qing Wang 1,2,3,4 , Fanjiang Xu 1,2,3,4∗ 1 Institute of Software Chinese Academy of Sciences, Beijing, China 2 Science & Technology on Integrated Information System Laboratory, Beijing, China 3 State Key Laboratory of Complex System Modeling and Simulation Technology, Beijing, China 4 University of Chinese Academy of Sciences, Beijing, China 5 Beijing University of Post and Telecommunications, Beijing, China arXiv :2608 . 16544v1 [ cs .MA] 17 Aug 2026 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 implementationspecific public skill changes into reusable, structured versionchange experience and adaptively fuses it with trajectoryderived 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. 1 Introduction Large language model agents increasingly rely on external skills as reusable units of task knowledge (Li et al. 2026; Ling, Zhong, and Huang 2026) . A skill can package instructions, scripts, resources, and configuration files into a portable bundle that an agent can invoke when solving a class of tasks (Anthropic 2025, 2026a) . However, a skill is rarely finished after its first release. As tools change, task distributions shift, and failures accumulate, skill authors revise instructions, patch scripts, add resources, etc. Skill evolution is part of how agent capabilities become reliable over time (Zhao et al. 2024; Wang et al. 2024) . Recent work on self-evolving agents has started to automate this process. A typical skill self-evolution loop executesan agent on tasks, collects trajectories, diagnoses failures, and revises the current skill accordingly (Zhao et al. 2024; Yang et al. 2026). This trajectory-driven paradigm grounds updates ∗Corresponding authors. Copyright © 2027, Association for the Advancement of Artificial Intelligence ([www.aaai.org](www.aaai.org)). All rights reserved. in task-specific execution evidence. However, its guidance is bounded by the quality and coverage of the trajectories collected in the current run (Mi et al. 2026). Noisy or low-quality trajectories may make it difficult to derive reliable guidance, while incomplete coverage of failure modes may cause the resulting updates to overfit to the limited observation. Consequently, trajectory-driven evolution may overlook broadly useful update knowledge that is not revealed by current executions (Huang et al. 2026) . A complementary source of evolution knowledge lies in public version histories. Software engineering research has mined historical patches to recover reusable edit and repair patterns (Bader et al. 2019; Koyuncu et al. 2020) . Public repositories similarly preserve how developers revise instructions, scripts, and other components during maintenance. These changes may encode update strategies that are not observed in the current ta

FAQs

What problem does VCE-Skill address in skill self-evolution?

It addresses the limitation of trajectory-driven evolution that depends on the quality and coverage of trajectories from current task runs, which can miss broadly useful update knowledge.

How does VCE-Skill use public skill version histories?

It distills raw, noisy, implementation-specific public version changes into structured version-change experience that can be reused across tasks.

How is the external version-change experience combined with trajectory evidence?

VCE-Skill uses an adaptive experience attention mechanism that selects relevant distilled experience while a base evolver independently generates a trajectory-derived self-proposal, then fuses them with attention weights.

Sep 20, 2026

Uploaded by Margaret

VCE-Skill - Enhancing Skill Self-Evolution with Version-Change Experience_Page 1

Open in WPS to view all 23 pages