SkillGate is a method that trains agents to select the correct skill from a large slate during an episode by separating credit signals for skill selection and execution. It addresses the problem of selector credit starvation, where traditional outcome-rewarded RL fails to give sufficient credit to the skill-naming tokens, especially in long-horizon tasks. Experiments on five benchmarks show that SkillGate improves a 9B policy’s success rate from 40.8% to 53.2%, reduces exposure to misleading candidates, and requires fewer skill reads.
By Qingyao Li, Wenxiang Jiao, Shuai Shao, Kangning Zhang, Yuan Lu, Yi Guo, Weiwen Liu, Weinan Zhang, Yong Yu
arXiv:2606. 14239v1 Announce Type: new Abstract: Agent skills are structured procedural packages that guide frozen LLM agents in specialized workflows.
By Haowen Gao, Haoran Chen, Can Wang, Shasha Guo, Liang Pang, Zhaoyang Liu, Huawei Shen, Xueqi Cheng
Agentic reinforcement learning (RL) enables large language models (LLMs) to learn from multi-turn interactions with verifiable outcomes. Textual skills can improve exploration by distilling prior inte...
arXiv:2610.09430v1 Announce Type: new
Abstract: Internalizing external skills changes a language agent's capabilities and, with them, the value of its remaining guidance: rules can become redundant,...
By Ling Li, Qiuyu Shen, Zheng Jiang, Qinwei Ma, Yuxuan Liu, Zhidong Deng
arXiv:2607. 21419v1 Announce Type: new Abstract: In long-horizon LLM agent reinforcement learning, weak policies often repeat similar failures, producing uninformative rollout trajectories and limiting effective policy optimization.
By Yipeng Shi, Zhipeng Ma, Yue Wang, Qitai Tan, Yang Li, Peng Chen, Zhengzhou Zhu
arXiv:2610.10426v1 Announce Type: new
Abstract: Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery. Exist...
By Jixuan Chen, Jiaxin Zhang, Qinyuan Ye, Yada Pruksachatkun, Haoxiang Zhang, Jingming Zhuo, Yifan Zhang, Yutong Dai, Juntao Tan, Xiangyu Peng, Silvio Savarese, Zeyuan Chen, Lianhui Qin, Chien-Sheng Wu
SkillEvoReg is a regularization framework designed to mitigate overfitting in language-model agents that evolve reusable external skills. It combines training-time skill dropout, complexity-aware local regularization, and causal counterexample validation to control skill-state growth and detect regressions. Applied across SkillOpt, SkillEvolBench, and ContinualSkillBench, it preserves downstream performance while improving transfer and later-stage evolution outcomes.
By Guanyu Nie, Fangzhou Zhu, Shixiong Kai, Xiongwei Han, Tao Zhong, Mingxuan Yuan
arXiv:2608. 05080v1 Announce Type: new Abstract: Critic-free group-based reinforcement learning has become a scalable approach for post-training large language models.
By Zheyuan Zhang, Manqing Mao, Hong Wang, Zhuoer Wang, Samson Koelle, Jie Yuan, Yanjun Lin, James Feng, Nikki Lijing Kuang, Yanfang Ye, Wei Niu
arXiv:2608. 06880v1 Announce Type: new Abstract: General-purpose skills promise reusable procedural knowledge for language agents, yet semantic relevance does not guarantee execution utility: a retrieved skill may encode assumptions that conflict with the current task, execution environment, or other retrieved skills.
By Qinfeng Li, Dalin He, Yuntai Bao, Ying Yang, Ruoxi Chen, Xinyan Yu, Lizhou Liang, Ge Su, Wenqi Zhang, Xuhong Zhang
arXiv:2606. 01619v1 Announce Type: new Abstract: Agentic reinforcement learning (RL) enables LLM agents to improve continuously from environment rewards, yet the resulting policies do not systematically accumulate reusable strategies that generalize across tasks.
By Zelin He, Haotian Lin, Boran Han, Wei Zhu, Haoyang Fang, Bernie Wang, Xuan Zhu, Runze Li, Matthew Reimherr
MInTRL (Minimal Intervention Reinforcement Learning) expands exploration in on-policy reinforcement learning by inserting sparse, local corrections into rollouts via a judge-intervention policy. These interventions replace erroneous suffixes and immediately return control to the main policy, allowing the agent to explore beyond its natural trajectory while maintaining on-policy data. The method uses a sequence-level advantage-regression objective, avoiding importance sampling, and demonstrates superior performance on math and code benchmarks compared to standard on-policy and off-policy baselines.
By Mingyu Chen, Yefan Tao, Gerald Friedland, Xuezhou Zhang, Chris Kong
The paper investigates how a higher‑level controller in a hierarchical latent reasoner decides whether to preserve or replace a lower‑level strategy. Experiments across different training seeds show that learned revision timing can produce varied policies, but none surpasses a fixed, forced timing policy on the same checkpoint. The study highlights that state‑dependent control can alter actions without improving task performance, and that timing decisions are order‑sensitive and largely captured by a strong fixed schedule.
By Ayushi Chadha