LLM Agent Training

LLM: Large Language Model

Follow this topic Ask about this topic

Momentum

54 papers in the last four weeks, up 145% on the four weeks before. 0.5% of all new papers.

Jul 13Week of Sep 28

Latest papers 354

All topics

EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning
Aug 6, 2026Zishan Xu, Zhiyuan Yao, Yuxin Chen +9LLM World Models Agentic RL
Shanghai Jiao Tong University · Zhejiang University · National University of Singapore +4
StepReflect: Structured UI Transition Reflection for Mobile GUI Agents
Aug 6, 2026Linqiang Guo, Wei Liu, Li Gu +3Mobile GUI Automation GUI Agents
Peter · Concordia University Montreal, Quebec, Canada
Recursive Synthesis for Long-Horizon Terminal Tasks
Aug 5, 2026Zhongzhi Li, Yucheng Shi, Zongxia Li +8Terminal Agents Synthetic Benchmark Generation
Tencent HY LLM Frontier · University of Georgia · University of Maryland, College Park +5
ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment
Aug 5, 2026Yijun Lu, Rui Ye, Jiajun Wang +4Credit Assignment in RL Step-Level Credit Assignment in RL
Shanghai Jiao Tong University
State2State: Environment-Derived Mid-Training for LLM Agents
Aug 5, 2026Xuanyu Lei, Yiqi Zhu, Chenliang Li +6LLM Agent Training Goal-Conditioned RL
Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China · Dept. of Comp. Sci. & Tech., Institute for AI, Tsinghua University, Beijing, China · Institute of Intelligent Computing, Alibaba Group
A-SR: Self-Evolving Agentic LLMs for Symbolic Regression via Hierarchical Coordination
Aug 5, 2026Wenxiao Zhao, Dong Liu, Kaiyi Xu +9Multi-Agent LLM Systems LLM Agent Training
Shanghai AI Laboratory · University of California at Los Angeles · Tongji University
Getting the Parameters Right: A Difficulty-Graded Benchmark and Probe-Guided Training for LLM Tool Calls
Aug 4, 2026Guoyao Yu, Xiaoqing Sun, Ziqi Huang +13Tool-Augmented Language Model Agents AI Agent Benchmarks
Qwen-CUA: Native Computer Use for (almost) Everything
Aug 3, 2026Dunjie Lu, Shuai Bai, Tianyi Bai +43Computer-Use Agent Benchmarks Computer-Use Agents
Trajectories That Segment Themselves: Agent-Declared Boundaries as a Training Unit
Aug 3, 2026Jingxi WeiLLM Agent Training Long-Horizon Agent Tasks
Look Ahead Before You Distill: Future Trajectory Validation of Teacher Guidance for Agentic On-Policy Distillation
Aug 3, 2026Chishui Chen, Yaoyou Fan, Te Sun +11LLM Agent Training On-Policy Distillation
Meituan LongCat Interaction · Peking University · Shanghai Jiao Tong University +6
CRISP: Critical Step Perception for Training Efficient Deep Search Agents
Aug 3, 2026Haosi Mo, Zihao Yan, Ruiqing Zhang +4LLM Agents LLM Agent Training
Baidu Inc., China · Institute of Computing and Intelligence, Harbin Institute of Technology, Shenzhen, China
SearchMaster: Grounded and Regulated Self-Play for Search Agents
Aug 3, 2026Wentao Tan, Qiong Cao, Jiaqi Wang +1Multi-Hop QA Tool-Augmented Language Model Agents
JD Future Academy, Beijing
SyncPlan: Long-Horizon LLM Coordination with Explicit Synchronization and Adaptive Correction
Aug 3, 2026Shen You, Xiaoming Zhu, Weining Weng +17Multi-Agent Coordination LLM Planning
1City University of Hong Kong · 2Tencent · 3Shenzhen University +1
Search-GRT: Guided Retrieval Training of Search Agents to Optimize for Complex Question Answering
Aug 2, 2026Aounon Kumar, Sudipta Paul, Vivek Kulkarni +2Multi-Hop QA Agentic RAG
AI Center-Mountain View, Samsung Electronics
Deep Research Pretraining via Predictive Navigation
Aug 1, 2026Jiang Zhou, Zhiyuan Fan, Xing Wu +3Deep Research Agents LLM Agent Training
Hunyuan Team, Tencent
Learning to Coordinate Symbolic Tools: LLM Agents for Verified Sum-of-Squares Certificates
Jul 31, 2026Bohan Chen, Shivam N. Patel, Richard Hoffmann +2LLM Agent Verification Tool-Using Agents
1California Institute of Technology · 2OpenAI
MAGA: Multi-Platform Self-Fusion of GUI Agents via Structured Action Distillation
Jul 31, 2026Hang Yan, Zhangxuan GU, Beitong Zhou +5Multi-Teacher Knowledge Distillation LLM Agent Training
1Xi’an Jiaotong University · 2Ant Group · 3Shanghai Jiao Tong University
Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI Agents
Jul 30, 2026Hanzhang Zhou, Panrong Tong, Xu Zhang +13Mobile GUI Automation Computer-Use Agents
MAI-UI Team Alibaba Token Hub, Alibaba Group
Echoverse: Deep, Evolving Environments for Training Computer-Use Agents at Scale
Jul 30, 2026Yash Pandya, Sahil Gupta, Sarthak Harne +10Computer-Use Agent Benchmarks Computer-Use Agents
Microsoft Research · Work done while at Microsoft Research.
SKILL-KD: Contrastive Skill Distillation for LLM Agents
Jul 30, 2026Qiming Shi, Yibo Dou, Jiawen Zhu +5LLM Agent Skill Learning LLM Agent Training
State Key Lab of CAD&CG, Zhejiang University · School of Software and Microelectronics, Peking University · School of Software Technology, Zhejiang University +1
ClawTrack: Towards Trace-Level Evaluation and Improvement of Real-World Autonomous Agents
Jul 30, 2026Xingjian Wu, Xuhang Zhu, Xingchen Liu +6AI Agent Reliability LLM Agent Evaluation
Meituan
Meta-Task: Turning Terminal Task Synthesis into a Terminal Task for Scalable Agent Training
Jul 30, 2026Zhihong Pan, Jiyuan He, Kai Zhang +5Synthetic Data Generation Terminal Agents
State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China · Meituan
MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis
Jul 29, 2026Yihao Chen, Shi Chang, Khaled Chawa +4Program Synthesis Automated Software Engineering
2Queen’s University · 1Huawei Canada · University of Manitoba
WikiLoop: Jointly Learning to Build and Navigate Agent-Native Wikis with Downstream Feedback
Jul 29, 2026Haoliang Ming, Feifei Li, Wenhui QueMulti-Hop QA KG Construction
WeChat, Tencent Inc., Beijing, China
Speculate While You Reason: Teaching Agents to Predict Their Next Tool Call via Joint Agent-Speculator RL
Jul 28, 2026Jiabao Ji, Yujian Liu, Li An +4Tool-Using Agents LLM Agent Training
University of California, Santa Barbara · Linkedin Inc
A Control System, a Dataset, and a Recipe for Making Frozen LLM Agents Learn a Domain
Jul 28, 2026Debjyoti PaulAgent Harness Optimization LLM Agent Training
Matryoshka Agent: Unfolding Sub-Agents for Long-Horizon Machine Learning Engineering
Jul 27, 2026Rushi Qiang, Changhao Li, Haotian Sun +3LLM Agent Orchestration LLM Agents
Georgia Institute of Technology
ACM: Agentic Context Management for Long Horizon Tasks
Jul 26, 2026Xiaochuan Li, Ryan Ming, Meng Chu +3LLM Agent Training Long-Horizon Agent Tasks
Carnegie Mellon University · Meta
AgentOmnia: Scaling Agentic Models for Full-Scenario Applications
Jul 25, 2026Hao Jiang, Gangtao Xin, Yingdi Huang +35Tool-Augmented Language Model Agents Agentic RL
Huawei Cloud Post-Training Team
SearchArt: Training Long-Horizon Search Agent with Scalable Synthetic and Verified Task
Jul 25, 2026Lang Mei, Xiaohan Yu, Chong Chen +27Synthetic Data Generation LLM Agents
Huawei Cloud Post-Training Team

Aug 6, 2026, cs.AIJ/K move · Enter open · S save

EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning

Zishan Xu, Zhiyuan Yao, Yuxin Chen, Yifu Guo+8

Shanghai Jiao Tong University · Zhejiang University · National University of Singapore+4

LLM World Models Agentic RL World Models LLM Agent Training RL for Tool Use

Open paper AskPDF arXiv

Training large language model agents for long-horizon tool use typically relies on interactions with real or synthesized executable environments, whose construction and verification are costly, or on external simulators that are difficult to ground. We introduce EnvACE, an agentic reinforcement learning method that replaces external environment interaction during training with world rehearsal. The policy alternates between acting and rehearsal: it first generates a tool call, then plays the role of the environment to produce the response induced by that action, and conditions subsequent decisions on the rehearsed response. Both roles are jointly optimized end-to-end using task-success rewards. Through world rehearsal, the policy internalizes the relationship between actions and their environment responses in its parameters, yielding an agent world model that directly supports decision making. Across BFCL-v4, tau^2-Bench, VitaBench, and FinMCP-Bench, EnvACE achieves strong and transferable performance, outperforming environment-scaling baselines in the overall evaluation. Controlled studies further show that world rehearsal consistently improves policy learning across model scales. At test time, the internalized world model enables private rehearsal before committed execution, yielding further gains under a moderate rehearsal budget without additional external interaction. Our findings establish world rehearsal as a new path toward scaling LLM agent training beyond the constraints of external environments. Our code is publicly available at https://github.com/Within-yao/EnvACE.

Momentum

Related topics

Latest papers 354