The Year in AI Papers
The year/Independent research

Paper 2605.06130

Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learning

Yaorui Shi, Yuxin Chen, Zhengxi Lu, Yuchun Miao, Shugui Liu, Qi GU, Xunliang Cai, Xiang Wang, An Zhang

Published
May 2026
Research lab
Independent
Citations
10
GitHub
108 stars

01 In brief

Summary

The paper introduces Skill1, a framework for training a single policy to co-evolve skill selection, utilization, and distillation in language model agents.

Unlike prior methods that optimize these capabilities in isolation or with separate reward sources, Skill1 derives all learning signals from a single task-outcome reward by decomposing it into a low-frequency trend (for selection) and high-frequency variation (for distillation).

The policy generates a query, re-ranks retrieved skills, executes the task conditioned on the selected skill, and distills new skills from successful trajectories.

Experiments on ALFWorld and WebShop show Skill1 achieves 97.5% success on ALFWorld, surpassing prior baselines like RetroAgent by 2.6 points.

Ablations confirm that removing any credit signal degrades all three capabilities, demonstrating their mutual dependence.

The framework also shows improved selection precision, library quality, and skill diversity compared to ablations.

Limitations include evaluation on only two text-based environments and a fixed library capacity of 5,000 skills.

02 From the paper

Abstract

A persistent skill library allows language model agents to reuse successful strategies across tasks. Maintaining such a library requires three coupled capabilities. The agent selects a relevant skill, utilizes it during execution, and distills new skills from experience. Existing methods optimize these capabilities in isolation or with separate reward sources, resulting in partial and conflicting evolution. We propose Skill1, a framework that trains a single policy to co-evolve skill selection, utilization, and distillation toward a shared task-outcome objective. The policy generates a query to search the skill library, re-ranks candidates to select one, solves the task conditioned on it, and distills a new skill from the trajectory. All learning derives from a single task-outcome signal. Its low-frequency trend credits selection and its high-frequency variation credits distillation. Experiments on ALFWorld and WebShop show that Skill1 outperforms prior skill-based and reinforcement learning baselines. Training dynamics confirm the co-evolution of the three capabilities, and ablations show that removing any credit signal degrades the evolution.