Long-Horizon LLM Agents

LLM: Large Language Model

Long-horizon LLM agents are systems designed to pursue goals through many connected actions, often while using tools and responding to changing information. Common approaches include hierarchical planning and subgoals, reinforcement learning with step-level or milestone rewards, persistent memory and context compression, and harnesses that verify actions and recover from errors. Researchers evaluate them in interactive settings such as ALFWorld, WebShop, ScienceWorld, WebArena, software-engineering benchmarks, and deep-research tasks. A central challenge is making performance reliable over time: errors and lost constraints can compound, while memory, verification, and extra computation can improve success but add cost or introduce new failure modes.

Latest papers

451 papers92 papers in the last four weeks, up 56% on the four weeks before. 0.5% of all new papers.

Don't Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation
Oct 29, 2026Zheng Zhu, Xiaozhuan Liang, Xi Chen +2Long-Horizon LLM Agents Memory Management for LLM Agents
The Chinese University of Hong Kong · Tencent · The Hong Kong University of Science and Technology
Spine: Compiling Subtask Trees into Structured Context for Long-Horizon Coding Agents
Oct 27, 2026Jiahong Xiang, Kunqiu Chen, Lingming Zhang +1Long-Horizon LLM Agents Software Engineering Agents
Southern University of Science and Technology, Shenzhen, China · University of Illinois at Urbana-Champaign, Urbana, IL, USA
AutoVend: Benchmarking LLM Agents on Long-Horizon Autonomous Vending Operations
Oct 27, 2026Qiming Shi, Yikun Li, Beidi Luan +5Long-Horizon LLM Agents LLM Agent Evaluation
StepFun · Zhejiang University · Xi’an Jiaotong-Liverpool University +1
MarkovSim: an LLM Agent Harness with Bounded Context for Long-Horizon Computational Simulation
Oct 26, 2026Yeui-Lung Lei, Wen-Bo Chen, Pan ZhangLLM Agent Harnesses Long-Horizon LLM Agents
College of Civil and Transportation Engineering, Shenzhen University, Shenzhen, 518060, China · Department of Civil and Environmental Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China.
Multiscale Policy Optimization for Long-Horizon LLM Agents
Oct 26, 2026Tao Ren, Hui Yang, Jia Deng +16Policy Optimization RL for LLM Agents
Peking University · IQuest Research · Tsinghua University +3
AgentBaton: Efficient Model Collaboration in Agentic Tasks
Oct 25, 2026Junlong Li, Jindong Jiang, Yonggan Fu +6LLM Agent Efficiency LLM Agent Orchestration
The Hong Kong University of Science and Technology · Project Leads · University of Illinois Urbana-Champaign
InteractiveWorld: A Verifiable Human-Centric Environment for Building Better Interactive MLLM Assistants
Oct 25, 2026Qinsi Wang, Wei Wen, Chong Zhou +7Multimodal Agent Evaluation Human-in-the-Loop AI
Meta · Duke University
FlexA-Mem: Budget-Adaptive Agentic Memory for LLM Agents
Oct 23, 2026Wujiang Xu, Jiaojiao Han, Xi Zhu +5Memory Management for LLM Agents Long-Term Conversational Memory
Rutgers University · Independent Researcher · Stanford University +2
LeafSkill: Efficient Leaf-Level Action Grounding for Recursive LLM Agents
Oct 23, 2026Shen Liu, Ze Zhou, Zhenguo Xu +3LLM Agent Skill Retrieval LLM Grounding
North China Institute of Computer System Engineering, Beijing, China · Beijing University of Aeronautics and Astronautics, Beijing, China · University of Science and Technology of China, Hefei, China +1
ReflectPO: Reflection-Guided Policy Optimization for Long-Horizon LLM Agents
Oct 22, 2026Rumei Li, Xunliang CaiLong-Horizon LLM Agents RL for LLM Agents
Meituan, China
HGM-Dual: Lineage-Aware Guided Agent Harness Optimization for Long-Horizon Planning
Oct 21, 2026Sudipta Paul, Pengrui Quan, Aounon Kumar +3Agent Harness Optimization LLM Agent Self-Improvement
Samsung Research America
Steering Black-Box Large Language Model Agents with Pitfall Anticipation World Model
Oct 21, 2026Jiongxiao Wang, Han Ding, Youzhi Luo +4LLM World Models LLM Agents
University of Wisconsin–Madison · AWS AI
Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition
Oct 20, 2026Wenjie Lou, Tianshuo Peng, Shiyang Feng +96Long-Horizon LLM Agents AI Agents for Scientific Discovery
OptionWorld: A Benchmark for Diagnosing Failures of LLMs as Long-Term Personal Assistants
Oct 20, 2026Ruoyu Zhang, Zhaocen Liu, Weiyi Tian +2LLM Agent Evaluation LLM Agents
University of British Columbia · University of Chicago · Independent Researcher +1
FALCA: Falsification-Aware Reinforcement Learning with Hierarchical Credit Assignment for Long-Horizon Deep Search Agents
Oct 20, 2026Zijun Dou, Kun Luo, Chaofan Li +2Long-Horizon LLM Agents RL for LLM Agents
BAAI · THU · CASIA
Harness Self-Evolution for Long-Horizon Terminal Agents
Oct 20, 2026Ruhan Wang, Yucheng Shi, Zongxia Li +6Agent Harness Optimization LLM Agent Reliability
Tencent HY LLM Frontier · Indiana University · University of Maryland, College Park +2
PORA: Learning When More Reasoning Pays Off for LLM Agents
Oct 19, 2026Bo Yan, Qi Zhang, Zhen Tan +2Adaptive Reasoning in Language Models Efficient Language Model Reasoning
University of Central Florida · Independent Researcher · Stevens Institute of Technology +1
M-Plan: Comprehensive Planning For LLM Agents with Memory Arrangement
Oct 19, 2026Ziyang Liu, Jiakang Yuan, Jiayuan Fan +1Multi-Agent LLM Systems Memory Management for LLM Agents
College of Future Information Technology, Fudan University, Shanghai, China · College of Intelligent Robotics and Advanced Manufacturing, Fudan University, Shanghai, China · Shanghai Innovation Institute, Shanghai, China
ClueFlow: Learning Long-Horizon Search through Attributed Research-State Transitions
Oct 19, 2026Yicheng Wang, Wentao Tan, Yidan Huang +2Long-Horizon LLM Agents RL for LLM Agents
Joy Future Academy
Let Them Sleep: Recursive Self-Pruning Memory for Long-Horizon LLM Agents
Oct 18, 2026Kargi Chauhan, Sadiba Nusrat Nur, Pratibha Revankar +2Memory Consolidation for LLM Agents LLM Agent Memory
University of California, Santa Cruz
Editing as Language: Teaching LLM Agents a Grammar for Native Context Editing
Oct 18, 2026Yunlong Deng, Yan Li, Xing Hu +4LLM Context Management LLM Agent Training
The Chinese University of Hong Kong · Meituan LongCat Team · Nanyang Technological University +2
LAMB: Latent Multi-Hypothesis Beliefs for Long-Horizon Large Language Agents
Oct 18, 2026Peien Ding, Hui Li, Yuhan Liu +1Long-Horizon LLM Agents Memory-Augmented Language Model Agents
School of Informatics, Xiamen University · MBZUAI
RL-Guided Temporal Localization for Dual-Channel Retrieval in Long-Horizon Agent Memory
Oct 16, 2026Dabin Sheng, Zhe Wu, Jinming Zhao +2Temporal IR Agentic Retrieval
Key Laboratory of Pervasive Computing, Ministry of Education, Department of Computer Science and Technology, Tsinghua University · Qiyuan Lab
KiranaBench: Can a Language Model Run a Neighbourhood Shop?
Oct 16, 2026Bhuvan Arora, Ankita Karmakar, Chaitanya Jain +10AI Agent Benchmarks Long-Horizon LLM Agents
BITS Pilani, Pilani Campus · University of Minnesota, Twin Cities · BITS Pilani, K. K. Birla Goa Campus
Long-Horizon RL Needs Muon
Oct 14, 2026Kai Ruan, Jinghao Lin, Tao Chen +3Long-Horizon RL Long-Horizon LLM Agents
Gaoling School of Artificial Intelligence, Renmin University of China · Duke University · Institute of Automation, Chinese Academy of Sciences +1
Explorer-Zero: Improving Self-Play for Long-Horizon Tasks with Decomposable Verifiable Rewards
Oct 14, 2026Tianyu Liu, Davide Belli, Bence Major +2Reward Decomposition Long-Horizon LLM Agents
ETH Zurich · Qualcomm AI Research‡
LAGO: Language Agents with Reusable Growing Options
Oct 14, 2026Yuran Wu, Mengkang Hu, Qionglin Qiu +5Long-Horizon LLM Agents LLM Agent Skill Learning
Shenzhen University · University of Hong Kong · Hunan University +3
New Model, Old Habits: Behavior Transfer under Model Swapping in Multi-Session LLM Agents
Oct 14, 2026Youngbin Choi, Sehyun Hwang, Suhyeon Jun +4LLM Agents Multi-LLM Collaboration
Graduate School of Artificial Intelligence, POSTECH · Meta · Department of Computer Science and Engineering, POSTECH
L3M: A Verifiable Harness and Runtime for Self-Extending and Self-Improving Agents
Oct 14, 2026Carson Eisenach, Robert Joseph George, Vincent Quenneville-Belair +2LLM Agent Harnesses Runtime Enforcement for AI Agents
Amazon
$τ$ -Emotion-Bench: Exposing the Interaction Fragility of LLM Agents through Real-World User Emotions
Oct 14, 2026Rishitosh Kumar Singh, Amir Saeidi, Shamanthak Hegde +2LLM Agent Evaluation Tool-Using Agents
Arizona State University

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

Don't Lose the Thread: Empowering Long-Horizon LLM Agents with Cognitive Resource Self-Allocation

Zheng Zhu, Xiaozhuan Liang, Xi Chen, Bei Yu+1

The Chinese University of Hong Kong · Tencent · The Hong Kong University of Science and Technology

Long-Horizon LLM Agents Memory Management for LLM Agents Context Compaction for LLM Agents RL for LLM Agents

Open paper AskPDF arXiv

Agents powered by large language models (LLMs) have demonstrated remarkable progress in solving complex reasoning tasks. However, LLM agents often falter on long-horizon tasks due to cognitive overload, as their working memory becomes cluttered with expanding and irrelevant information, which dilutes their attention and hinders effective planning and reasoning. To mitigate this challenge, we introduce Cognitive Resource Self-Allocation (CORAL), a novel reasoning paradigm that empowers agents to proactively optimize their context. Implemented as an agent-callable working memory management toolset, CORAL allows an agent to checkpoint its task progress and verified facts, and adaptively initiate a new problem-solving episode by resetting cluttered context while preserving only the checkpointed information, enabling it to resume reasoning from a clean state. We further enhance the agent's checkpointing capabilities using a Multi-episode Agentic Reinforced Policy Optimization algorithm. On the GAIA benchmark, CORAL with an 8B model achieves 42.8% average accuracy, with particularly strong gains on the most challenging long-horizon tasks (Level 2 and Level 3). Crucially, these gains come at a substantially lower inference cost: CORAL consumes 6-7

\times

fewer input tokens than a ReAct agent. It also generalizes to multi-hop Knowledge QA benchmarks and to the considerably harder BrowseComp, where it improves a strong proprietary agent by 8.0 points.