OpenSkill: Open-World Self-Evolution for LLM Agents
Summary
OpenSkill is a framework for LLM agents to self-evolve skills and verification signals from open-world resources without target-task supervision, achieving high performance across benchmarks.
View Cached Full Text
Cached at: 06/08/26, 03:30 AM
Paper page - OpenSkill: Open-World Self-Evolution for LLM Agents
Source: https://huggingface.co/papers/2606.06741
Abstract
OpenSkill enables self-evolving agents to develop skills and verification signals from scratch using open-world resources without target-task supervision, achieving high automated performance across benchmarks.
Self-evolving agentsrequires adaptation after deployment, but existing approaches assume a usable learning loop, such as curated skills, successful trajectories, or verifier signals. Realopen-world deployments may provide none of these, offering only a task prompt. In this work, we study open-world self-evolution, where an agent must build both its skills and its ownverification signalsfrom scratch, using open-world resources but no target-task supervision. We propose OpenSkill, a framework that bootstraps this loop: it acquiresgrounded knowledgeand verification anchors from documentation, repositories, and the web, synthesizes them intotransferable skills, and refines those skills againstself-built virtual tasksgrounded in the anchors rather than in target answers. The open world thus supplies both the knowledge to be learned and a supervision-independent practice environment, with target-task supervision reserved for final evaluation. Across three benchmarks and two target agents, OpenSkill attains the bestautomated pass ratewhile satisfying the no-supervision constraint. Analysis shows its skills transfer across models without model-specific adaptation, and its self-built verifier aligns with ground-truth outcomes despite never accessing them.
View arXiv pageView PDFProject pageAdd to collection
Get this paper in your agent:
hf papers read 2606\.06741
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2606.06741 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2606.06741 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.06741 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
OpenSkillEval: Automatically Auditing the Open Skill Ecosystem for LLM Agents
OpenSkillEval is an automatic evaluation framework for auditing open-source skills used by LLM agents across multiple downstream tasks. Using over 600 dynamically generated tasks and 30 skills, the authors find that skill availability does not guarantee effective usage and that benefits depend heavily on the model and framework.
Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills
The paper introduces Skill Self-Play (Skill-SP), a co-evolutionary framework that uses a proposer, solver, and skill controller to bridge structured verification and open-ended exploration, improving LLM performance on tool-use and reasoning benchmarks.
@dair_ai: // MetaSkill-Evolve // Great paper on self-improving agents. Most self-improving agents rewrite what the agent does and…
MetaSkill-Evolve introduces a recursive two-timescale framework for LLM agents to evolve both task skills and the improvement procedure itself, achieving notable accuracy gains on OfficeQA, SealQA, and ALFWorld benchmarks.
Training LLM Agents for Spontaneous, Reward-Free Self-Evolution via World Knowledge Exploration
This paper proposes a method to train LLM agents with intrinsic meta-evolution capabilities, enabling spontaneous self-improvement without external rewards at inference time. Applied to Qwen3-30B and Seed-OSS-36B, the approach yields a 20% performance boost on web navigation benchmarks, with a 14B model outperforming Gemini-2.5-Flash.
SkillOpt: Executive Strategy for Self-Evolving Agent Skills
SkillOpt introduces a systematic text-space optimizer for agent skills that trains skills as external agent state with stable updates and zero deployment inference overhead, achieving superior performance across multiple benchmarks and execution environments.