WikiSkill Gains Increase With Model Scale
Tu Vu reports WikiSkill gains grow across Qwen model sizes instead of shrinking.
Tu Vu posted that WikiSkill gains grow across Qwen model sizes instead of diminishing. He links the pattern to chain-of-thought prompting and states the method co-evolves agent skills with a persistent wiki knowledge base that stores experiences. Andrew Drozdov reposted the observation on X. The performance claims originate from Tu Vu's posts and linked thread and have not been independently verified.
One of my favorite patterns in ML/AI: some methods get *more* useful as models scale. @_jasonwei showed this strikingly with chain-of-thought. We observe a similar trend with WikiSkill across Qwen model sizes: rather than diminishing, its gains increase with model scale!
🚨 Excited to share our new work on WikiSkill (led by the awesome @LiyanTang4)! We propose a method that co-evolves agent skills and a wiki that integrates agent experience into a growing knowledge base. WikiSkill substantially outperforms current SoTA skill evolution methods…
Reactions
2 postsTV 
Tu Vu@tuvllmsOne of my favorite patterns in ML/AI: some methods get *more* useful as models scale. @_jasonwei showed this strikingly with chain-of-thought. We observe a similar trend with WikiSkill across Qwen model sizes: rather than diminishing, its gains increase with model scale!
- likes: 20
- replies: 1
- bookmarks: 6
- reposts: 3
TV Tu Vu@tuvllms🚨 Excited to share our new work on WikiSkill (led by the awesome @LiyanTang4)! We propose a method that co-evolves agent skills and a wiki that integrates agent experience into a growing knowledge base. WikiSkill substantially outperforms current SoTA skill evolution methods 👇
- likes: 21
- replies: 3
- bookmarks: 7
- reposts: 6
