VCE-Skill: Enhancing Skill Self-Evolution with Version-Change Experience
3.00T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2608.16544.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryA research paper proposing VCE-Skill, which improves agent skill self-evolution by distilling public skill version-change histories into structured experience and adaptively fusing it with execution-trajectory-derived proposals. Experiments report 3.20–4.98 point score gains and stronger cross-model skill transfer.
Why it mattersProposes a concrete method for enriching agent skill libraries by mining public version histories — a previously untapped source — with measured gains over trajectory-only baselines.
Cited by
No citations on record.
