SkillRise: Agentic Reinforcement Learning for Cross-Task Skill Evolution

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Zhiyuan Yao, Yuxin Chen, Zhengxi Lu, Zishan Xu, Yueqing Sun, Yifu Guo, Yuquan Lu, Zhengzhou Cai, Kangning Zhang, Zhuowen Han, Zi-Han Wang, Ziang Ye, Qi Gu, Xunliang Cai, Weiwen Liu, Yongliang Shen
• Venue: arXiv: cs.LG
• Published: 2026-07-29
• Collected: 2026-07-30
• URL: http://arxiv.org/abs/2607.26784v1
• PDF: https://arxiv.org/pdf/2607.26784v1
• DOI: UNKNOWN

Plain-Language Brief

The paper introduces SkillRise, a reinforcement learning framework that allows large language model agents to learn and reuse skills across related yet distinct tasks. SkillRise improves Pass@1 performance on ALFWorld, WebShop, and ScienceWorld compared to other methods, scales effectively at test time across tasks, and reduces runtime overhead of skill learning pipelines.

What the Document Says

• SkillRise is a unified reinforcement learning framework for learning skills across related tasks.
• It organizes related instances into progressively challenging sequences.
• 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 show that SkillRise achieves the strongest Pass@1 performance among compared methods on ALFWorld, WebShop, and ScienceWorld.
• Gains over the strongest baseline range from 2.3 to 8.5 percentage points.
• Trained across distinct tasks, but its learned curation policy remains effective for repeated attempts on the same task.
• 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 retains strong performance while reducing runtime overhead of skill learning pipelines with multiple stages.

Key Concepts Mentioned

• Large language model agents
• Reinforcement learning
• Skill evolution
• Task solving
• Curation policy
• Pass@1 performance
• ALFWorld
• WebShop
• ScienceWorld

Claims Made by the Document

• SkillRise improves Pass@1 performance on ALFWorld, WebShop, and ScienceWorld compared to other methods.
• Trained across distinct tasks, but its learned curation policy remains effective for repeated attempts on the same task.
• Scaling at test time across tasks: performance improves with longer sequences of related tasks even when each task is attempted only once.

Evidence Used

• Results from experiments on ALFWorld, WebShop, and ScienceWorld
• Analysis of scaling trends at test time across tasks

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

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.