LLM Agent Skill Learning

LLM: Large Language Model

This field asks how language-model agents can turn experience, demonstrations, external guides, or feedback into reusable procedures for tasks such as tool use, web navigation, coding, and embodied control. Common approaches distill skills from successful and failed trajectories into text, executable programs, or hierarchical libraries, then retrieve, compose, verify, and revise them; other methods train agents to internalize skills through supervised fine-tuning or reinforcement learning. Work is evaluated on interactive benchmarks such as ALFWorld, WebShop, AppWorld, and coding tasks, with attention to success, transfer, token cost, and consistency. A central challenge is that skills can help when relevant but cause regressions when stale, poorly selected, or applied outside their scope, making reliable validation, lifecycle management, and safety essential.

Latest papers

401 papers71 papers in the last four weeks, up 39% on the four weeks before. 0.4% of all new papers.

Compile Away Uncertainty: Self-Evolving Agent Skills via Programmatic Compilation
Oct 29, 2026Chengyu Yang, Yunxiang Zheng, Zongzheng Li +6LLM Agent Skill Learning Self-Evolving Agents
East China Normal University · China University of Geoscience
Before the Next Rollout: Diagnosing Skill Utility Shifts from Policy Updates
Oct 29, 2026Yifan Wang, Zhenyi Gai, Yiwen Zhao +5LLM Agent Skill Evaluation LLM Agent Skill Learning
Fudan University · Pyra
SkillMem: Compact External Skill Memory for Latent Multi-Agent Reasoning
Oct 27, 2026Yuxuan Zhu, Mohammad Mohammadi Amiri, Pin-Yu Chen +1Procedural Memory for LLM Agents Memory-Augmented Language Model Agents
Rensselaer Polytechnic Institute · IBM Research
Skill Dreaming: Evolving Repository Skills through Pull Request Streams
Oct 26, 2026Xinyuan Song, Hasibul Haque, Jay Somasundaram +3Automated Software Engineering LLM Agent Skill Learning
Causal Dynamics Lab, USA
Back to Basics: A Skill-Centric Approach to Generalist Web Agents
Oct 26, 2026Minh Pham-Dinh, Md Farhan Ishmam, Hernan Yupanqui +2Web Agents LLM Agent Skill Learning
University of Utah · Islamic University of Technology
Hierarchical Skill Evolution for Coding Agent
Oct 26, 2026Jun Gao, Yun Peng, Changhai Zhou +10LLM Agent Skill Learning Software Engineering Agents
Skill $^{\infty}$ : Recursive Self-Improvement for Agent Skill Optimization
Oct 26, 2026Wangcheng Tao, Zhanglu Yan, Han Wu +1LLM Agent Self-Improvement LLM Agent Skill Learning
National University of Singapore · City University of Hong Kong
MultiEvoSkill: Structure-Aware Multi-Agent Skill Evolution for Generalizable Skill Artifacts
Oct 25, 2026Kounianhua Du, Jianxing Liu, Ruoshui Shan +4LLM Agent Skill Learning Large Language Model-Guided Optimization
Shanghai Jiao Tong University Shanghai, China · Purdue · Tencent
$Δτ$ -Bench: Pushing Agents Toward a Moving Capability
Oct 25, 2026Shuofei Qiao, Xuanguang Pan, Yijun Chen +6AI Agent Benchmarks LLM Agent Evaluation
Zhejiang University · University College London
Lifelong Context Adaptation through Bidirectional Memory-Skill Learning
Oct 23, 2026Tianyi Xu, Huazheng WangMemory-Augmented Language Model Agents Continual Learning for LLM Agents
Oregon State University
Phronix: The Capability-Centric Learning Paradigm for Agents
Oct 22, 2026Hariharan Devarajan, Neeraj Rajesh, Aniruddha Marathe +3LLM Agent Skill Learning LLM Agent Harnesses
Lawrence Livermore National Laboratory · Illinois Institute of Technology · University of Tennessee, Knoxville
Charts as Epistemic Actions for Experience-Driven Agentic Data Analysis
Oct 21, 2026Xinzhe Wang, Xudong Weng, Zang Li +1Data Analysis Agents LLM Agent Skill Learning
Tencent · Renmin University of China
Robust Soft Skill Adaptation via Behavioral Localization in Tool-Using Agents
Oct 21, 2026Zijian Jiang, Xilu WangLLM Agent Skill Learning Prompt Tuning
School of Computer Science University of Surrey Guildford, Surrey GU2 7XH, UK
D-Skill: Distilling Teacher Knowledge through Agent Skills
Oct 20, 2026Shuo Yuan, Dacheng Li, Joseph E. GonzalezSkill Distillation Knowledge Distillation
University of California, Berkeley
Automating ML for Science: Can Frontier Agents Climb Scientific Hills in the Wild?
Oct 20, 2026Ming Zhong, Stacy Li, Nicholas Carlini +1AI Agents for Scientific Discovery Scientific Agent Benchmarks
Anthropic Fellows Program · Anthropic
ChartEvo: Self-Evolving Agent Harnesses for Long-Tail Chart Reasoning
Oct 20, 2026Qi Wang, Peijie Wang, Jun Cao +4LLM Agent Self-Improvement LLM Agent Harnesses
State Key Laboratory of Multimodal Artificial Intelligence Systems, CASIA · School of Artificial Intelligence, University of Chinese Academy of Sciences · New Laboratory of Pattern Recognition, CASIA +1
Which Agent's Skill Should Change? Group-Relative Credit Assignment for Evolving Multi-Agent Teams
Oct 19, 2026Yanan Ma, Yang Wang, David Le +6LLM Agent Skill Learning Multi-Agent LLM Systems
Mayo Clinic · University of Manchester · Alibaba Token Hub, Alibaba Group +1
Procedural Exposure Bias in Skill-Augmented Language Model Agents
Oct 19, 2026Qing Yang, Yuqi Xiong, Pengcheng Huang +5LLM Agents Cognitive Bias in Language Models
School of Computer Science and Engineering, Northeastern University, Shenyang, China · Alibaba Group, Hangzhou, China
EvoSkill v2: Fully Agentic Skill Evolution from Benchmark Adaptation to a Stateful Evolver
Oct 19, 2026Dastin Huang, Baran Nama, Salaheddin AlzubiLLM Agent Skill Learning Self-Evolving Agents
Sentient Labs
Standing on the Shoulders of Giants: Enhancing LLM Agents for Automated Fact-Checking with Human Expert Skills
Oct 19, 2026Xinwen Chen, Haorui He, Yupeng LiAutomated Fact-Checking LLM Agent Skill Learning
Department of Interactive Media, Hong Kong Baptist University · School of Computing and Data Science, The University of Hong Kong · AI and Social Good Lab, AI Media Centre, Hong Kong Baptist University
Trackable Skill Library Lifecycle Management for Agentic Reinforcement Learning
Oct 18, 2026Xi Zhu, Yuanlin Duan, Changyi Yang +4LLM Agent Skill Learning Agentic RL
Rutgers University · Carnegie Mellon University · TikTok Inc.
AtomSD: Procedural Atom-Guided Self-Distillation for LLM Agents
Oct 17, 2026Bohan Yu, Pengfei Cao, Kai Tian +8LLM Agent Skill Learning RL for LLM Agents
School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences · The Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences · School of Artificial Intelligence, University of Chinese Academy of Sciences +1
Selm: Self-Evolving Library Management for Multi-Skill Agent Optimization
Oct 17, 2026Mengxi Hu, Yuxuan Hu, Jinbo Su +6LLM Agent Skill Learning LLM Agents
School of Information, Renmin University of China, Beijing, China · Key Laboratory of Data Engineering and Knowledge Engineering, Beijing, China · Beijing Academy of Artificial Intelligence, Beijing, China +2
DrugHarness: An Evolvable Agent Harness for Computational Drug Discovery
Oct 16, 2026Yuxing Chen, Zhisheng Zhang, Cheng Tang +20LLM Agent Harnesses Drug Discovery
Shanghai Artificial Intelligence Laboratory · China Pharmaceutical University · University of the Chinese Academy of Sciences +7
From Trajectories to Differentiated Skills: Cognitive Differentiation for Self-Improving Agents
Oct 16, 2026Yewei Zhao, Derong Xu, Wei Chen +6LLM Agent Skill Learning Self-Improving Agents
University of Science and Technology of China · City University of Hong Kong · Meituan
Darwin-Claw: Environment-Adaptive Experience Consolidation for Recursively Self-Evolving Agents
Oct 16, 2026Mingze Yin, Xiaohan Wang, Dian Li +4Self-Evolving Agents Continual Learning for LLM Agents
Zhejiang University · Tencent · Zhongguancun Academy
R $^2$ Orch: Recursive Self-Improvement for Agent Orchestration via Self-Play within a Skill-Internalization Loop
Oct 15, 2026Mingda Zhang, Yun Huang, Qiang Huang +7LLM Agent Orchestration Self-Improving Agents
The Chinese University of Hong Kong, Shenzhen, China · Fudan University, China · University of Illinois Urbana-Champaign, USA +2
RSIFlow: Adaptive Self-Improvement Workflows with an Evolving Meta-Agent
Oct 15, 2026Huan Zhang, Mingju Chen, Dongxu Zhou +4LLM Agent Self-Improvement Self-Evolving Agents
Beijing Advanced Innovation Center for Future Blockchain and Privacy Computing, School of Artificial Intelligence, Beihang University · Tsinghua University · Beijing Academy of Artificial Intelligence (BAAI)
MetaExp: Learning to Self-Improve through Experience via Verifiable Cross-Task Environment Scaling
Oct 15, 2026Renhao Li, Xinjie Shen, Yang Su +4LLM Agent Self-Improvement LLM Agents
University of Macau · Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences · Georgia Institute of Technology +1
FedSkillGit: Federated Skill Evolution via Version Control
Oct 15, 2026Boxun Lan, Jiamei Li, Weijian Li +5LLM Agent Skill Learning Federated Learning
East China Normal University, Shanghai, China · Shandong University, Jinan, China

Oct 29, 2026J/K move · Enter open · S save

Compile Away Uncertainty: Self-Evolving Agent Skills via Programmatic Compilation

Chengyu Yang, Yunxiang Zheng, Zongzheng Li, Jiamei Li+5

East China Normal University · China University of Geoscience

LLM Agent Skill Learning Self-Evolving Agents LLM-Based Program Synthesis Compiler-Guided Program Synthesis

Open paper AskPDF arXiv

Self-evolving agent skills enable AI agents to accumulate and refine reusable procedural knowledge from interaction experience without updating model parameters. However, since such skills are typically represented as textual documents and executed through LLM reasoning, even previously learned procedures must be repeatedly interpreted and regenerated, introducing avoidable uncertainty into every execution. In contrast, program execution is deterministic and independently verifiable. Intuitively, transforming well-specified operations into executable programs can reduce reliance on LLM reasoning and, consequently, reduce execution uncertainty. Motivated by this intuition, we introduce SKillCompiler, a self-evolving skill framework that integrates programmatic compilation into the skill evolution process. Specifically, SkillCompiler adopts an executability-aware skill compilation strategy that decomposes skills into fine-grained operations, analyzes their executability, and constructs hybrid skill skeletons that retain LLM reasoning for context-dependent operations while compiling well-specified ones into validated programs. To avoid redundant program synthesis and enable executable knowledge to accumulate across evolving skills, we further design a program asset reuse mechanism that organizes validated programs as reusable assets, retrieves existing implementations whenever applicable, and synthesizes and verifies new programs only when necessary. Extensive experiments on five benchmarks and six model families demonstrate that SkillCompiler consistently outperforms state-of-the-art methods, achieving up to 32.33% improvement in task performance.