OpenSkill: Self-Evolving LLM Agents With No Task Supervision
OpenSkill lets agents build skills and their own verifiers from the open web, hitting 43.6% on SkillsBench (+8.9 over the best baseline) with zero target-task answers.
Institution
Public research university in Chicago, USA; its data mining and AI group (led by Philip S. Yu) is prominent in agent and graph learning research.
OpenSkill lets agents build skills and their own verifiers from the open web, hitting 43.6% on SkillsBench (+8.9 over the best baseline) with zero target-task answers.