EASE: Behavior-Adaptive Skill Curation for Self-Evolving Agents

2026-09-29

AI Summary

This study addresses the systematic cross-executor performance degradation in existing skill curation frameworks caused by their failure to model downstream executor behavior. To overcome this limitation, we propose EASE, a pioneering behavior-adaptive skill curation paradigm that enables a single curator to dynamically adapt across multiple executors without fine-tuning. Specifically, EASE conditions a shared curator via online behavioral profiling and incorporates retrieval-augmented, behavior-aware temporal credit assignment to facilitate self-evolving skill library management. Experimental evaluations on benchmarks such as ALFWorld demonstrate that EASE consistently outperforms existing baselines. Notably, it reduces skill library size by 34.5%–41.0%, improves retrieval efficiency by 36.3%–38.7%, and decreases inference token consumption by 9.1%–14.5%, highlighting its effectiveness in building scalable, executor-agnostic skill repositories.

Abstract

Agent skills provide a lightweight mechanism for self-evolving agents to accumulate reusable procedural knowledge without updating model parameters. However, existing learned skill curators typically optimize curation without explicitly modeling downstream executor behavior. We show that this can cause systematic cross-executor degradation: curators trained with different executors perform best when paired with their own training executor, indicating that effective skill curation is executor-dependent. We formulate behavior-adaptive skill curation and introduce EASE, a framework that learns a single curator that adapts its decisions to different executor behaviors. EASE maintains an online behavioral profile of recent execution patterns and conditions the curator on this profile, the current trajectory, and retrieved skills to add, modify, or remove skills from an evolving repository. We train the shared curator jointly across multiple frozen executors with reinforcement learning, using retrieval-aware and behavior-aware temporal attribution to focus optimization on curation actions with observable downstream influence. Across ALFWorld, ScienceWorld, and WebShop, with executors ranging from Qwen3-8B/32B and GPT-OSS-120B to unseen Kimi K2.6, DeepSeek V4 Flash, and Gemini 3.5 Flash, EASE outperforms strong skill- and memory-based baselines without per-executor finetuning. EASE also maintains 34.5--41.0% fewer skills, improves skill retrieval by 36.3--38.7% and measured edit utility by 51.8--60.0%, and reduces deployment-time inference tokens by 9.1--14.5%. These results establish behavior-adaptive skill curation as an effective principle for building self-evolving agents.

Problem

• self-evolving agents
• skill curation
• cross-executor degradation
• behavior-adaptive
• reusable procedural knowledge

Innovation

• Behavior-Adaptive Skill Curation
• Self-Evolving Agents
• Reinforcement Learning
• Temporal Attribution
• Online Behavioral Profile