SkillCycle: Co-Evolving Agent Policies and Skill Banks

Key points

• Internalizing external skills changes a language agent's capabilities and, with them, the value of its remaining guidance: rules can become redundant, misleading, or insufficient for newly encountered decisions.1
• This creates a coupled problem of learning from skills and adapting the skills that supervise further learning.1
• SkillCycle alternates between two phases: policy learning with a fixed skill bank and router, and rule revision with a frozen policy.1
• These results show that continually revising skill guidance as the agent's capabilities change helps transform external skills into policy capabilities that require no skill inputs at inference.1

Sources (1)

• [1]SkillCycle: Co-Evolving Agent Policies and Skill Banks
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 04:33 AM

  We introduce SkillCycle, a framework for co-evolving agent policies and skill banks through a feedback loop between skill internalization and rule revision.

  Internalizing external skills changes a language agent's capabilities and, with them, the value of its remaining guidance: rules can become redundant, misleading, or insufficient for newly encountered decisions.

Extractive summary: sentences quoted from the sources.