Saved in:
Bibliographic Details
Main Authors: Lin, Hongxiang, Kuai, Zhirui, Xue, Erpeng, Wang, Lei
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.27899
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917538272641024
author Lin, Hongxiang
Kuai, Zhirui
Xue, Erpeng
Wang, Lei
author_facet Lin, Hongxiang
Kuai, Zhirui
Xue, Erpeng
Wang, Lei
contents Structured skill prompts improve exploration in long-horizon agentic reinforcement learning (RL). Skill-augmented RL methods retain external skills at inference, while skill-internalization RL methods withdraw them during training to enable autonomous performance. However, existing internalization approaches only use skill-helpfulness contrast for curriculum control, leaving the policy update unchanged and unable to distinguish skill-dependent from autonomous success. We propose SkillC, a framework based on Contrastive Skill Credit Assignment (CSCA) that converts this contrast into a direct learning signal for internalization. \textsc{SkillC} samples paired skill-injected and skill-free rollouts for tasks from active skill types within the same policy update, and injects their task-level contrast into optimization via a dual-stream advantage estimator that preserves global ranking while applying a one-sided correction toward skill-free success. A smoothed validation-level signal further drives an adaptive curriculum over attribution strength, rollout allocation, and monotonic active-set pruning. Experiments on ALFWorld and WebShop show that, without runtime skill access, SkillC surpasses the strongest prior skill-internalization RL baseline by 5.5\% and 4.4\%, respectively, while remaining competitive with skill-augmented RL methods.
format Preprint
id arxiv_https___arxiv_org_abs_2605_27899
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SKILLC: Learning Autonomous Skill Internalization in LLM Agents via Contrastive Credit Assignment
Lin, Hongxiang
Kuai, Zhirui
Xue, Erpeng
Wang, Lei
Artificial Intelligence
Structured skill prompts improve exploration in long-horizon agentic reinforcement learning (RL). Skill-augmented RL methods retain external skills at inference, while skill-internalization RL methods withdraw them during training to enable autonomous performance. However, existing internalization approaches only use skill-helpfulness contrast for curriculum control, leaving the policy update unchanged and unable to distinguish skill-dependent from autonomous success. We propose SkillC, a framework based on Contrastive Skill Credit Assignment (CSCA) that converts this contrast into a direct learning signal for internalization. \textsc{SkillC} samples paired skill-injected and skill-free rollouts for tasks from active skill types within the same policy update, and injects their task-level contrast into optimization via a dual-stream advantage estimator that preserves global ranking while applying a one-sided correction toward skill-free success. A smoothed validation-level signal further drives an adaptive curriculum over attribution strength, rollout allocation, and monotonic active-set pruning. Experiments on ALFWorld and WebShop show that, without runtime skill access, SkillC surpasses the strongest prior skill-internalization RL baseline by 5.5\% and 4.4\%, respectively, while remaining competitive with skill-augmented RL methods.
title SKILLC: Learning Autonomous Skill Internalization in LLM Agents via Contrastive Credit Assignment
topic Artificial Intelligence
url https://arxiv.org/abs/2605.27899