Hugging Face Daily Papers · · 4 min read

SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

We introduce SkillRise, a unified RL framework where a single policy alternates between solving tasks and curating an evolving skill document across related tasks, with decoupled credit assignment enabling cross-task skill transfer and test-time scaling.</p>\n","updatedAt":"2026-07-30T03:30:41.468Z","author":{"_id":"676127cf11b19ea602bb202a","avatarUrl":"/avatars/dfd802a24bd63e509728159ebb1769f6.svg","fullname":"Zhengxi Lu","name":"LZXzju","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":10,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8925411701202393},"editors":["LZXzju"],"editorAvatarUrls":["/avatars/dfd802a24bd63e509728159ebb1769f6.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.26784","authors":[{"_id":"6a6ac4a84463a8a84bdc3fc1","name":"Zhiyuan Yao","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc2","name":"Yuxin Chen","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc3","name":"Zhengxi Lu","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc4","name":"Zishan Xu","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc5","name":"Yueqing Sun","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc6","name":"Yifu Guo","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc7","name":"Yuquan Lu","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc8","name":"Zhengzhou Cai","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fc9","name":"Kangning Zhang","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fca","name":"Zhuowen Han","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fcb","name":"Zi-Han Wang","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fcc","name":"Ziang Ye","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fcd","name":"Qi Gu","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fce","name":"Xunliang Cai","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fcf","name":"Weiwen Liu","hidden":false},{"_id":"6a6ac4a84463a8a84bdc3fd0","name":"Yongliang Shen","hidden":false}],"publishedAt":"2026-07-29T00:00:00.000Z","submittedOnDailyAt":"2026-07-30T00:00:00.000Z","title":"SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution","submittedOnDailyBy":{"_id":"676127cf11b19ea602bb202a","avatarUrl":"/avatars/dfd802a24bd63e509728159ebb1769f6.svg","isPro":false,"fullname":"Zhengxi Lu","user":"LZXzju","type":"user","name":"LZXzju"},"summary":"Large language model agents often encounter related yet distinct tasks that share reusable solution patterns. Yet standard agentic reinforcement learning treats tasks as independent episodes, while existing approaches to skill learning either focus on repeated attempts of one task or use pipelines with multiple stages that entangle extraction, retrieval, and execution. We introduce SkillRise, a unified reinforcement learning framework for learning skills across tasks. SkillRise organizes related instances into progressively challenging sequences and uses a single policy to alternate between task solving and curating an evolving skill document passed directly to the next task. Decoupled credit assignment across tasks supervises solving with the current task outcome and curation with discounted downstream outcomes. Experiments on ALFWorld, WebShop, and ScienceWorld show that SkillRise achieves the strongest Pass@1 performance among the compared methods, with gains over the strongest baseline ranging from 2.3 to 8.5 percentage points. Although trained across distinct tasks, its learned curation policy remains effective for repeated attempts on the same task. Further analysis reveals scaling at test time across tasks: performance improves with longer sequences of related tasks even when each task is attempted only once. This trend suggests that SkillRise reuses transferable skills across tasks rather than benefiting from repeated sampling of the same task. SkillRise further retains strong performance while substantially reducing the runtime overhead of skill learning pipelines with multiple stages. Together, these results provide a simple and efficient training paradigm for LLM agents to extract, refine, and reuse transferable skills across tasks.","upvotes":13,"discussionId":"6a6ac4a94463a8a84bdc3fd1","githubRepo":"https://github.com/Within-yao/SkillRise","githubRepoAddedBy":"user","githubStars":3},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6708edcae69f6e30a816af9f","avatarUrl":"/avatars/c4daa9b0cb2f4bb2a7db0e78b22034cb.svg","isPro":false,"fullname":"Yao","user":"distant-yuan","type":"user"},{"_id":"676127cf11b19ea602bb202a","avatarUrl":"/avatars/dfd802a24bd63e509728159ebb1769f6.svg","isPro":false,"fullname":"Zhengxi Lu","user":"LZXzju","type":"user"},{"_id":"699edcd4796be456dc1dfaf7","avatarUrl":"/avatars/7d9112dbd6e723b7c945f542ac619782.svg","isPro":false,"fullname":"杨紫瑜","user":"ELIJAHHI3","type":"user"},{"_id":"69b7a148b71e93b5b8d4cb51","avatarUrl":"/avatars/544e2fc71b81e85df396e7d03e462138.svg","isPro":false,"fullname":"Zi-Han Wang","user":"zzzzhw","type":"user"},{"_id":"65d2c31b053a863f53c9892e","avatarUrl":"/avatars/f0a7b76ed4910813699bcb8d576e1e95.svg","isPro":false,"fullname":"yza","user":"ziangLeaf","type":"user"},{"_id":"68a2fb20b3ad3d518527c49d","avatarUrl":"/avatars/cb37f5b2a98cbbea6f9726552a8f60e2.svg","isPro":false,"fullname":"zq","user":"capturee","type":"user"},{"_id":"63b6def76fca9d2a1902fa14","avatarUrl":"/avatars/c7f2487450ea954e2bca4fc5a6db8eb3.svg","isPro":false,"fullname":"张康宁","user":"zhangkangning","type":"user"},{"_id":"643910dbabdc6ce5351e4eb5","avatarUrl":"/avatars/92ec189cd4325b4d85fdfcd59f1ff1e3.svg","isPro":false,"fullname":"Yueqing Sun","user":"leqing","type":"user"},{"_id":"6603e7b679ed63106e974434","avatarUrl":"/avatars/4a7de974260da697ebab1136ee070c9f.svg","isPro":false,"fullname":"Yuquan Lu","user":"Eric1019","type":"user"},{"_id":"674572a99543fbaf3c63f35b","avatarUrl":"/avatars/6c891450c2ceeb7b034556548afc772d.svg","isPro":false,"fullname":"蔡正舟","user":"conctsai","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"5e1058e9fcf41d740b69966d","avatarUrl":"/avatars/ce74839ba871f2b54313a670a233ba82.svg","isPro":false,"fullname":"Yongliang Shen","user":"tricktreat","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.26784.md","query":{}}">
Papers
arxiv:2607.26784

SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution

Published on Jul 29
· Submitted by
Zhengxi Lu
on Jul 30
Authors:
,

Abstract

Large language model agents often encounter related yet distinct tasks that share reusable solution patterns. Yet standard agentic reinforcement learning treats tasks as independent episodes, while existing approaches to skill learning either focus on repeated attempts of one task or use pipelines with multiple stages that entangle extraction, retrieval, and execution. We introduce SkillRise, a unified reinforcement learning framework for learning skills across tasks. SkillRise organizes related instances into progressively challenging sequences and uses a single policy to alternate between task solving and curating an evolving skill document passed directly to the next task. Decoupled credit assignment across tasks supervises solving with the current task outcome and curation with discounted downstream outcomes. Experiments on ALFWorld, WebShop, and ScienceWorld show that SkillRise achieves the strongest Pass@1 performance among the compared methods, with gains over the strongest baseline ranging from 2.3 to 8.5 percentage points. Although trained across distinct tasks, its learned curation policy remains effective for repeated attempts on the same task. Further analysis reveals scaling at test time across tasks: performance improves with longer sequences of related tasks even when each task is attempted only once. This trend suggests that SkillRise reuses transferable skills across tasks rather than benefiting from repeated sampling of the same task. SkillRise further retains strong performance while substantially reducing the runtime overhead of skill learning pipelines with multiple stages. Together, these results provide a simple and efficient training paradigm for LLM agents to extract, refine, and reuse transferable skills across tasks.

Community

Paper submitter about 5 hours ago

We introduce SkillRise, a unified RL framework where a single policy alternates between solving tasks and curating an evolving skill document across related tasks, with decoupled credit assignment enabling cross-task skill transfer and test-time scaling.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2607.26784
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper

No model linking this paper

Cite arxiv.org/abs/2607.26784 in a model README.md to link it from this page.

Datasets citing this paper

No dataset linking this paper

Cite arxiv.org/abs/2607.26784 in a dataset README.md to link it from this page.

Spaces citing this paper

No Space linking this paper

Cite arxiv.org/abs/2607.26784 in a Space README.md to link it from this page.

Collections including this paper

No Collection including this paper

Add this paper to a collection to link it from this page.

Discussion (0)

Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.

Sign in →

No comments yet. Sign in and be the first to say something.

More from Hugging Face Daily Papers