Agent Systems with Harness Engineering

May 25, 2026 ยท View on GitHub

A Curated Reading List and Structured Roadmap

paper openreview status references last-update license

A curated list of papers and resources on agent systems with harness engineering, based on our paper:
"Agent Systems with Harness Engineering".

Timeline of agent systems with harness engineering

Timeline of agent systems with harness engineering.

Tip: If you have any questions for our paper, please send an email to txy20010310@163.com.

Table of Contents

1. Evolution of Harness Engineering

Three layers in the evolution of harness architecture

Three layers in the evolution of harness architecture.

1.1. Action Interface: Connecting Model to Environment

  1. Quine: Realizing LLM Agents as Native POSIX Processes. [Paper] arXiv 2026
  2. ceLLMate: Sandboxing Browser AI Agents. [Paper] arXiv 2025
  3. Executable Code Actions Elicit Better LLM Agents. [Paper] ICML 2024
  4. DeepAgent: A General Reasoning Agent with Scalable Toolsets. [Paper] arXiv 2025
  5. ReAct: Synergizing Reasoning and Acting in Language Models. [Paper] [Code] ICLR 2023
  6. Toolformer: Language Models Can Teach Themselves to Use Tools. [Paper] NeurIPS 2023
  7. Tool learning with large language models: a survey. [Paper] FCS 2025
  8. AIOS: LLM Agent Operating System. [Paper] COLM 2025

1.2. Workflow Infrastructure: Orchestrating Persistent Workspaces

  1. SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering. [Paper] [Code] NeurIPS 2024
  2. Engineering Pitfalls in AI Coding Tools: An Empirical Study of Bugs in Claude Code, Codex, and Gemini CLI. [Paper] arXiv 2026
  3. Building Effective AI Coding Agents for the Terminal: Scaffolding, Harness, Context Engineering, and Lessons Learned. [Paper] arXiv 2026
  4. SWE-Master: Unleashing the Potential of Software Engineering Agents via Post-Training. [Paper] arXiv 2026
  5. VeRO: An Evaluation Harness for Agents to Optimize Agents. [Paper] ICML 2026
  6. OpenHands: An Open Platform for AI Software Developers as Generalist Agents. [Paper] ICLR 2025
  7. Claude Code Overview. [Docs] Docs 2026

1.3. User-Centric Persistence: Continuity Across Sessions and Channels

  1. Toward Personalized LLM-Powered Agents: Foundations, Evaluation, and Future Directions. [Paper] arXiv 2026
  2. Enabling Personalized Long-term Interactions in LLM-based Agents through Persistent Memory and User Profiles. [Paper] arXiv 2025
  3. User Preference Modeling for Conversational LLM Agents: Weak Rewards from Retrieval-Augmented Interaction. [Paper] arXiv 2026
  4. From Human Memory to AI Memory: A Survey on Memory Mechanisms in the Era of LLMs. [Paper] arXiv 2025
  5. Persistent Assistant: Seamless Everyday AI Interactions via Intent Grounding and Multimodal Feedback. [Paper] CHI 2025
  6. Introducing OpenClaw. [Blog] Blog 2026

2. The Design of the Harness

2.1. Agent Workflow

  1. HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face. [Paper] [Code] NeurIPS 2023
  2. OpenManus: An Open-source Framework for Building General AI Agents. [Code] Repo 2025
  3. Unrolling the Codex agent loop. [Blog] Blog 2026
  4. OpenClaw. [Code] [Website] Repo 2026
  5. Hermes Agent. [Website] Website 2026

2.1.1. Environment Perception

Observation Extraction
  1. Mind2Web: Towards a Generalist Agent for the Web. [Paper] [Code] NeurIPS 2023
  2. WebArena: A Realistic Web Environment for Building Autonomous Agents. [Paper] [Code] ICLR 2024
  3. WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models. [Paper] [Code] ACL 2024
  4. VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks. [Paper] [Code] ACL 2024
  5. GPT-4V(ision) is a Generalist Web Agent, if Grounded. [Paper] ICML 2024
  6. SeeClick: Harnessing GUI Grounding for Advanced Visual GUI Agents. [Paper] ACL 2024
  7. OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments. [Paper] [Code] NeurIPS 2024
  8. DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning. [Paper] [Code] NeurIPS 2024
  9. AgentOccam: A Simple Yet Strong Baseline for LLM-Based Web Agents. [Paper] ICLR 2025
  10. Agent S: An Open Agentic Framework that Uses Computers Like a Human. [Paper] [Code] ICLR 2025 Workshop
  11. AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents. [Paper] [Code] ICLR 2025
  12. GUI-Actor: Coordinate-Free Visual Grounding for GUI Agents. [Paper] [Code] NeurIPS 2025
  13. VAGEN: Reinforcing World Model Reasoning for Multi-Turn VLM Agents. [Paper] NeurIPS 2025
State Representation
  1. ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs. [Paper] [Code] ICLR 2024
  2. WebArena: A Realistic Web Environment for Building Autonomous Agents. [Paper] [Code] ICLR 2024
  3. SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering. [Paper] [Code] NeurIPS 2024
  4. Do LLMs Build World Representations? Probing Through the Lens of State Abstraction. [Paper] [Code] NeurIPS 2024
  5. AgentOccam: A Simple Yet Strong Baseline for LLM-Based Web Agents. [Paper] ICLR 2025
  6. AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials. [Paper] [Code] ICLR 2025
  7. GRAPHGPT-O: Synergistic Multimodal Comprehension and Generation on Graphs. [Paper] [Code] CVPR 2025
  8. The Complexity Trap: Simple Observation Masking Is as Efficient as LLM Summarization for Agent Context Management. [Paper] arXiv 2025
  9. ESCA: Contextualizing Embodied Agents via Scene-Graph Generation. [Paper] [Code] NeurIPS 2025
  10. Evaluating LLM Planning in Partially Observable Environments via Observation Representations and Action Sequences. [Paper] NeurIPS 2025 Workshop
  11. DynaWeb: Model-Based Reinforcement Learning of Web Agents. [Paper] arXiv 2026
  12. Building a C compiler with a team of parallel Claudes. [Blog] Blog 2026

2.1.2. Task Planning

Task Decomposition
  1. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. [Paper] [Code] NeurIPS 2022
  2. Least-to-Most Prompting Enables Complex Reasoning in Large Language Models. [Paper] ICLR 2023
  3. Tree of Thoughts: Deliberate Problem Solving with Large Language Models. [Paper] [Code] NeurIPS 2023
  4. Graph of Thoughts: Solving Elaborate Problems with Large Language Models. [Paper] [Code] AAAI 2024
  5. Agentic Reasoning: A Streamlined Framework for Enhancing LLM Reasoning with Agentic Tools. [Paper] ACL 2025
  6. Agentic Reasoning for Large Language Models. [Paper] arXiv 2026
  7. Building a C compiler with a team of parallel Claudes. [Blog] Blog 2026
Plan Generation
  1. AvaTaR: Optimizing LLM Agents for Tool Usage via Contrastive Reasoning. [Paper] [Code] NeurIPS 2024
  2. AFlow: Automating Agentic Workflow Generation. [Paper] [Code] ICLR 2025
  3. GAP: Graph-Based Agent Planning with Parallel Tool Use and Reinforcement Learning. [Paper] [Code] arXiv 2025

2.1.3. Action Execution

Tool Invocation
  1. Toolformer: Language Models Can Teach Themselves to Use Tools. [Paper] NeurIPS 2023
  2. Code as Policies: Language Model Programs for Embodied Control. [Paper] [Code] ICRA 2023
  3. SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering. [Paper] [Code] NeurIPS 2024
  4. AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents. [Paper] [Code] ACL 2024
  5. Code execution with MCP: building more efficient AI agents. [Blog] Blog 2025
  6. Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning. [Paper] arXiv 2025
  7. Scaling Agentic Reinforcement Learning for Tool-Integrated Reasoning in VLMs. [Paper] arXiv 2025
  8. Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors. [Paper] arXiv 2026
Environment Interaction
  1. ReAct: Synergizing Reasoning and Acting in Language Models. [Paper] [Code] ICLR 2023
  2. SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering. [Paper] [Code] NeurIPS 2024
  3. Grounding Multimodal Large Language Models in Actions. [Paper] NeurIPS 2024
  4. ceLLMate: Sandboxing Browser AI Agents. [Paper] arXiv 2025
  5. Agent-Fence: Mapping Security Vulnerabilities Across Deep Research Agents. [Paper] arXiv 2026
  6. Quine: Realizing LLM Agents as Native POSIX Processes. [Paper] arXiv 2026

2.2. Memory Systems

2.2.1. Short-term Memory

Working Memory
  1. ReAct: Synergizing Reasoning and Acting in Language Models. [Paper] [Code] ICLR 2023
  2. ReadAgent: A Human-Inspired Reading Agent with Gist Memory of Very Long Contexts. [Paper] [Code] ICML 2024
  3. Agent Workflow Memory. [Paper] [Code] ICML 2025
  4. Understanding Software Engineering Agents: A Study of Thought-Action-Result Trajectories. [Paper] ASE 2025
  5. Memo: Training Memory-Efficient Embodied Agents with Reinforcement Learning. [Paper] [Code] NeurIPS 2025
  6. Managing Context on the Claude Developer Platform. [Blog] Blog 2025
  7. Graph-based Agent Memory: Taxonomy, Techniques, and Applications. [Paper] arXiv 2026
Conversational Memory
  1. MemGPT: Towards LLMs as Operating Systems. [Paper] [Code] arXiv 2023
  2. Zep: A Temporal Knowledge Graph Architecture for Agent Memory. [Paper] arXiv 2025
  3. From Isolated Conversations to Hierarchical Schemas: Dynamic Tree Memory Representation for LLMs. [Paper] ICLR 2025
  4. SeCom: On Memory Construction and Retrieval for Personalized Conversational Agents. [Paper] ICLR 2025
  5. Hello Again! LLM-powered Personalized Agent for Long-term Dialogue. [Paper] NAACL 2025
  6. SGMem: Sentence Graph Memory for Long-Term Conversational Agents. [Paper] arXiv 2025
  7. Managing Context on the Claude Developer Platform. [Blog] Blog 2025
  8. Agentic Memory: Learning Unified Long-Term and Short-Term Memory Management for Large Language Model Agents. [Paper] arXiv 2026
  9. From Single to Multi-Granularity: Toward Long-Term Memory Association and Selection of Conversational Agents. [Paper] ICLR 2026
  10. FluxMem: Adaptive Hierarchical Memory for Streaming Video Understanding. [Paper] CVPR 2026

2.2.2. Long-term Memory

  1. Memory Overview. [Docs] Docs 2026
  2. LangMem. [Docs] Docs 2026
Structured Memory
  1. Generative Agents: Interactive Simulacra of Human Behavior. [Paper] [Code] UIST 2023
  2. Optimus-1: Hybrid Multimodal Memory Empowered Agents Excel in Long-Horizon Tasks. [Paper] [Code] NeurIPS 2024
  3. A-Mem: Agentic Memory for LLM Agents. [Paper] [Code] NeurIPS 2025
  4. Bridging Intuitive Associations and Deliberate Recall: Empowering LLM Personal Assistant with Graph-Structured Long-term Memory. [Paper] Findings of ACL 2025
  5. CarMem: Enhancing Long-Term Memory in LLM Voice Assistants through Category-Bounding. [Paper] COLING 2025
  6. Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory. [Paper] [Code] ECAI 2025
  7. G-Memory: Tracing Hierarchical Memory for Multi-Agent Systems. [Paper] [Code] NeurIPS 2025
  8. Graph-Native Cognitive Memory for AI Agents: Formal Belief Revision Semantics for Versioned Memory Architectures. [Paper] arXiv 2026
Unstructured Memory
  1. MemGPT: Towards LLMs as Operating Systems. [Paper] [Code] arXiv 2023
  2. Reflexion: language agents with verbal reinforcement learning. [Paper] [Code] NeurIPS 2023
  3. MemoryBank: Enhancing Large Language Models with Long-Term Memory. [Paper] [Code] AAAI 2024
  4. Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning. [Paper] [Code] arXiv 2025
  5. MOOM: Maintenance, Organization and Optimization of Memory in Ultra-Long Role-Playing Dialogues. [Paper] arXiv 2025
  6. O-Mem: Omni Memory System for Personalized, Long Horizon, Self-Evolving Agents. [Paper] arXiv 2025
  7. General Agentic Memory Via Deep Research. [Paper] arXiv 2025
  8. Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory. [Paper] [Code] arXiv 2025
  9. MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents. [Paper] [Code] ICLR 2026
  10. MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent. [Paper] [Code] ICLR 2026
  11. RetroAgent: From Solving to Evolving via Retrospective Dual Intrinsic Feedback. [Paper] arXiv 2026
  12. Agentic Memory: Learning Unified Long-Term and Short-Term Memory Management for Large Language Model Agents. [Paper] arXiv 2026
  13. Mem0: Universal Memory Layer for AI Agents. [Code] Repo 2026
  14. OpenMemory: AI Memory MCP Server for Coding Agents. [Website] Web 2026
  15. Memory - Docs by LangChain. [Docs] Docs 2026
  16. Context Engineering: Memory, Compaction, and Tool Clearing. [Docs] Docs 2026

2.3. Skill Libraries

  1. SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks. [Paper] [Code] arXiv 2026
  2. Memory - Docs by LangChain. [Docs] Docs 2026
  3. SkillX: Automatically Constructing Skill Knowledge Bases for Agents. [Paper] arXiv 2026

2.3.1. Skill Acquisition

Learning from Demonstration
  1. Toolformer: Language Models Can Teach Themselves to Use Tools. [Paper] NeurIPS 2023
  2. Large Language Models as Tool Makers. [Paper] [Code] ICLR 2024
  3. Gorilla: Large Language Model Connected with Massive APIs. [Paper] NeurIPS 2024
  4. Voyager: An Open-Ended Embodied Agent with Large Language Models. [Paper] [Code] TMLR 2024
  5. Synatra: Turning Indirect Knowledge into Direct Demonstrations for Digital Agents at Scale. [Paper] NeurIPS 2024
  6. ToolACE: Winning the Points of LLM Function Calling. [Paper] [Code] ICLR 2025
  7. SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skills. [Paper] arXiv 2025
  8. Agent Skill Acquisition for Large Language Models via CycleQD. [Paper] [Code] arXiv 2025
  9. Inducing Programmatic Skills for Agentic Tasks. [Paper] COLM 2025
  10. SoK: Agentic Skills - Beyond Tool Use in LLM Agents. [Paper] arXiv 2026
  11. Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forward. [Paper] arXiv 2026
  12. SkillCraft: Can LLM Agents Learn to Use Tools Skillfully? [Paper] [Code] arXiv 2026
  13. Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills. [Paper] arXiv 2026
  14. Go-Browse: Training Web Agents with Structured Exploration. [Paper] ICLR 2026
  15. ContractSkill: Repairable Contract-Based Skills for Multimodal Web Agents. [Paper] arXiv 2026
Learning from Experience
  1. Hierarchical Reinforcement Learning: A Survey and Open Research Challenges. [Paper] MAKE 2022
  2. ExpeL: LLM Agents Are Experiential Learners. [Paper] [Code] AAAI 2024
  3. OS-Copilot: Towards Generalist Computer Agents with Self-Improvement. [Paper] [Code] ICLR 2024
  4. SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience. [Paper] [Code] arXiv 2025
  5. Leveraging Skills from Unlabeled Prior Data for Efficient Online Exploration. [Paper] [Code] ICML 2025
  6. Reinforcement Learning for Self-Improving Agent with Skill Library. [Paper] [Code] arXiv 2025
  7. Toward Training Superintelligent Software Agents through Self-Play SWE-RL. [Paper] [Code] ICML 2026
  8. Evolving Programmatic Skill Networks. [Paper] arXiv 2026
  9. SkillRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning. [Paper] [Code] ICLR 2026 Workshop
  10. Tool-R0: Self-Evolving LLM Agents for Tool-Learning from Zero Data. [Paper] arXiv 2026
  11. AutoSkill: Experience-Driven Lifelong Learning via Skill Self-Evolution. [Paper] arXiv 2026
  12. Memento-Skills: Let Agents Design Agents. [Paper] arXiv 2026
  13. ELITE: Experiential Learning and Intent-Aware Transfer for Self-improving Embodied Agents. [Paper] arXiv 2026
  14. SKILL0: In-Context Agentic Reinforcement Learning for Skill Internalization. [Paper] [Code] arXiv 2026
  15. Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents. [Paper] [Project] arXiv 2026
Learning from External Resources
  1. Equipping Agents for the Real World with Agent Skills. [Blog] Blog 2025
  2. Agent Skills. [Docs] Docs 2025
  3. Skills in OpenAI API. [Docs] Docs 2026
  4. SKILLFOUNDRY: Building Self-Evolving Agent Skill Libraries from Heterogeneous Scientific Resources. [Paper] arXiv 2026

2.3.2. Skill Management

Skill Representation
  1. Voyager: An Open-Ended Embodied Agent with Large Language Models. [Paper] [Code] TMLR 2024
  2. Inducing Programmatic Skills for Agentic Tasks. [Paper] COLM 2025
  3. ToolGen: Unified Tool Retrieval and Calling via Generation. [Paper] [Code] ICLR 2025
  4. Equipping Agents for the Real World with Agent Skills. [Blog] Blog 2025
  5. SoK: Agentic Skills - Beyond Tool Use in LLM Agents. [Paper] arXiv 2026
  6. Evolving Programmatic Skill Networks. [Paper] arXiv 2026
  7. CUA-Skill: Develop Skills for Computer Using Agent. [Paper] [Code] arXiv 2026
  8. Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills. [Paper] arXiv 2026
  9. Memento-Skills: Let Agents Design Agents. [Paper] arXiv 2026
  10. SkillFlow: Scalable and Efficient Agent Skill Retrieval System. [Paper] arXiv 2026
  11. Graph of Skills: Dependency-Aware Structural Retrieval for Massive Agent Skills. [Paper] arXiv 2026
  12. SkillRouter: Skill Routing for LLM Agents at Scale. [Paper] arXiv 2026
  13. Skills in OpenAI API. [Docs] Docs 2026
  14. Agent Skills. [Docs] Docs 2026
Skill Retrieval
  1. Voyager: An Open-Ended Embodied Agent with Large Language Models. [Paper] [Code] TMLR 2024
  2. SRSA: Skill Retrieval and Adaptation for Robotic Assembly Tasks. [Paper] ICLR 2025
  3. EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecycle. [Paper] [Code] arXiv 2025
  4. Memento-Skills: Let Agents Design Agents. [Paper] arXiv 2026
  5. IntentCUA: Learning Intent-level Representations for Skill Abstraction and Multi-Agent Planning in Computer-Use Agents. [Paper] arXiv 2026
  6. GraphSkill: Documentation-Guided Hierarchical Retrieval-Augmented Coding for Complex Graph Reasoning. [Paper] arXiv 2026
  7. SkillRouter: Skill Routing for LLM Agents at Scale. [Paper] arXiv 2026
  8. SkillFlow: Scalable and Efficient Agent Skill Retrieval System. [Paper] arXiv 2026
  9. Graph of Skills: Dependency-Aware Structural Retrieval for Massive Agent Skills. [Paper] arXiv 2026
  10. WebXSkill: Skill Learning for Autonomous Web Agents. [Paper] arXiv 2026
  11. Skill Retrieval Augmentation for Agentic AI. [Paper] arXiv 2026

2.3.3. Skill Maintenance

Library Curation
  1. Using Skills to Accelerate OSS Maintenance. [Blog] Blog 2026
  2. Shell + Skills + Compaction: Tips for Long-Running Agents that Do Real Work. [Blog] Blog 2026
  3. SkillRouter: Skill Routing for LLM Agents at Scale. [Paper] arXiv 2026
  4. Graph of Skills: Dependency-Aware Structural Retrieval for Massive Agent Skills. [Paper] arXiv 2026
Skill Governance
  1. Equipping Agents for the Real World with Agent Skills. [Blog] Blog 2025
  2. Skills in OpenAI API. [Docs] Docs 2026
  3. Testing Agent Skills Systematically with Evals. [Blog] Blog 2026
  4. Towards Secure Agent Skills: Architecture, Threat Taxonomy, and Security Analysis. [Paper] arXiv 2026

2.4. Multi-agent Orchestration

2.4.1. Coordination Architectures

Centralized Architectures
  1. ChatDev: Communicative Agents for Software Development. [Paper] [Code] ACL 2024
  2. MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework. [Paper] [Code] ICLR 2024
  3. AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversations. [Paper] [Code] COLM 2024
  4. AutoAgents: A Framework for Automatic Agent Generation. [Paper] [Code] IJCAI 2024
  5. Scaling Large Language Model-based Multi-Agent Collaboration. [Paper] [Code] ICLR 2025
  6. AutoAgent: A Fully-Automated and Zero-Code Framework for LLM Agents. [Paper] [Code] arXiv 2025
  7. Agent S2: A Compositional Generalist-Specialist Framework for Computer Use Agents. [Paper] [Code] COLM 2025
  8. MegaAgent: A Large-Scale Autonomous LLM-based Multi-Agent System Without Predefined SOPs. [Paper] [Code] Findings of ACL 2025
  9. Multi-Agent Collaboration via Evolving Orchestration. [Paper] [Code] NeurIPS 2025
Decentralized Architectures
  1. CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society. [Paper] [Code] NeurIPS 2023
  2. AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors in Agents. [Paper] [Code] arXiv 2023
  3. ProAgent: Building Proactive Cooperative Agents with Large Language Models. [Paper] AAAI 2024
  4. A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration. [Paper] [Code] COLM 2024
  5. CONSENSAGENT: Towards Efficient and Effective Consensus in Multi-Agent LLM Interactions Through Sycophancy Mitigation. [Paper] [Code] Findings of ACL 2025
  6. AgentOrchestra: Orchestrating Multi-Agent Intelligence with the Tool-Environment-Agent(TEA) Protocol. [Paper] arXiv 2025
  7. LLM-Driven Multi-Agent Architectures for Intelligent Self-Organizing Networks. [Paper] IEEE Network 2025
  8. Building a C Compiler with a Team of Parallel Claudes. [Blog] Blog 2026

2.4.2. Communication Mechanisms

Debate-based Methods
  1. Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate. [Paper] [Code] EMNLP 2024
  2. Should we be going MAD? A Look at Multi-Agent Debate Strategies for LLMs. [Paper] ICML 2024
  3. MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework. [Paper] [Code] ICLR 2024
  4. RoleLLM: Benchmarking, Eliciting, and Enhancing Role-Playing Abilities of Large Language Models. [Paper] [Code] Findings of ACL 2024
  5. Debate-to-Write: A Persona-Driven Multi-Agent Framework for Diverse Argument Generation. [Paper] [Code] COLING 2025
  6. Table-Critic: A Multi-Agent Framework for Collaborative Criticism and Refinement in Table Reasoning. [Paper] ACL 2025
  7. Reinforce LLM Reasoning through Multi-Agent Reflection. [Paper] ICML 2025
  8. CONSENSAGENT: Towards Efficient and Effective Consensus in Multi-Agent LLM Interactions Through Sycophancy Mitigation. [Paper] [Code] Findings of ACL 2025
  9. Player-Coach Teamwork: Multi-agent Collaboration for Improving LLM Reasoning. [Paper] NeurIPS 2025 Workshop
Collaboration-based Methods
  1. ChatDev: Communicative Agents for Software Development. [Paper] [Code] ACL 2024
  2. MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework. [Paper] [Code] ICLR 2024
  3. GPTSwarm: Language Agents as Optimizable Graphs. [Paper] ICML 2024
  4. Long-Horizon Planning for Multi-Agent Robots in Partially Observable Environments. [Paper] [Code] NeurIPS 2024
  5. MapCoder: Multi-Agent Code Generation for Competitive Problem Solving. [Paper] [Code] ACL 2024
  6. Flow-of-Action: SOP Enhanced LLM-Based Multi-Agent System for Root Cause Analysis. [Paper] WWW 2025
  7. MAPoRL: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning. [Paper] [Code] ACL 2025
  8. MARFT: Multi-Agent Reinforcement Fine-Tuning. [Paper] [Code] arXiv 2025
  9. Multi-Agent Collaboration via Evolving Orchestration. [Paper] [Code] NeurIPS 2025
  10. AgentOrchestra: Orchestrating Multi-Agent Intelligence with the Tool-Environment-Agent(TEA) Protocol. [Paper] arXiv 2025
  11. Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL. [Paper] [Code] arXiv 2025
  12. Multi-Agent Evolve: LLM Self-Improve through Co-evolution. [Paper] [Code] arXiv 2025
  13. InstructFlow: Adaptive Symbolic Constraint-Guided Code Generation for Long-Horizon Planning. [Paper] [Code] NeurIPS 2025
  14. How and When to Build Multi-Agent Systems. [Blog] Blog 2025
  15. Don't Build Multi-Agents. [Blog] Blog 2025
  16. Toward Autonomous Long-Horizon Engineering for ML Research. [Paper] arXiv 2026

3. Model Adaptation for Harness

3.1. Context Engineering

3.1.1. Context Design

Prompt Engineering
  1. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. [Paper] [Code] NeurIPS 2022
  2. Large Language Models are Zero-Shot Reasoners. [Paper] NeurIPS 2022
  3. Tree of Thoughts: Deliberate Problem Solving with Large Language Models. [Paper] [Code] NeurIPS 2023
  4. Graph of Thoughts: Solving Elaborate Problems with Large Language Models. [Paper] [Code] AAAI 2024
  5. Role play with large language models. [Paper] Nature 2023
  6. CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society. [Paper] [Code] NeurIPS 2023
  7. Generative Agents: Interactive Simulacra of Human Behavior. [Paper] [Code] UIST 2023
  8. CoSER: A Comprehensive Literary Dataset and Framework for Training and Evaluating LLM Role-Playing and Persona Simulation. [Paper] [Code] ICML 2025
  9. Large Language Models Are Human-Level Prompt Engineers. [Paper] ICLR 2023
  10. Making Pre-trained Language Models Better Few-shot Learners. [Paper] ACL 2021
  11. Self-Refine: Iterative Refinement with Self-Feedback. [Paper] [Code] NeurIPS 2023
  12. Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution. [Paper] ICML 2024
Context Retrieval
  1. Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions. [Paper] ACL 2023
  2. ReAct: Synergizing Reasoning and Acting in Language Models. [Paper] [Code] ICLR 2023
  3. REPLUG: Retrieval-Augmented Black-Box Language Models. [Paper] [Code] NAACL 2024
  4. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection. [Paper] [Code] ICLR 2024
  5. CRAG: Corrective Retrieval Augmented Generation. [Paper] [Code] arXiv 2024
  6. RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval. [Paper] [Code] ICLR 2024
  7. Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity. [Paper] [Code] NAACL 2024
  8. HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models. [Paper] NeurIPS 2024
  9. Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph. [Paper] ICLR 2024
  10. DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines. [Paper] ICLR 2024
  11. StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization. [Paper] ICLR 2025
  12. Search-o1: Agentic Search-Enhanced Large Reasoning Models. [Paper] [Code] EMNLP 2025
  13. Deeprag: Thinking to retrieve step by step for large language models. [Paper] [Code] arXiv 2025
  14. Model Context Protocol (MCP): Landscape, Security Threats, and Future Research Directions. [Paper] [Code] arXiv 2025
  15. MA-RAG: Multi-Agent Retrieval-Augmented Generation via Collaborative Chain-of-Thought Reasoning. [Paper] arXiv 2025
  16. A-RAG: Scaling Agentic Retrieval-Augmented Generation via Hierarchical Retrieval Interfaces. [Paper] arXiv 2026

3.1.2. Context Management

Context Processing
  1. Lost in the Middle: How Language Models Use Long Contexts. [Paper] [Code] TACL 2024
  2. LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models. [Paper] EMNLP 2023
  3. Compressing Context to Enhance Inference Efficiency of Large Language Models. [Paper] [Code] EMNLP 2023
  4. LLM Maybe LongLM: Self-Extend LLM Context Window Without Tuning. [Paper] [Code] ICML 2024
  5. ACON: Optimizing Context Compression for Long-horizon LLM Agents. [Paper] [Code] arXiv 2025
  6. Reducing Cost of LLM Agents with Trajectory Reduction. [Paper] arXiv 2025
  7. AgentFold: Long-Horizon Web Agents with Proactive Context Management. [Paper] arXiv 2025
  8. SWE-Pruner: Self-Adaptive Context Pruning for Coding Agents. [Paper] arXiv 2026
  9. CEDAR: Context Engineering for Agentic Data Science. [Paper] [Code] arXiv 2026
  10. IterResearch: Rethinking Long-Horizon Agents with Interaction Scaling. [Paper] ICLR 2026
Context Updating
  1. Reflexion: language agents with verbal reinforcement learning. [Paper] [Code] NeurIPS 2023
  2. MemGPT: Towards LLMs as Operating Systems. [Paper] [Code] arXiv 2023
  3. Efficient Streaming Language Models with Attention Sinks. [Paper] [Code] ICLR 2024
  4. MemoryBank: Enhancing Large Language Models with Long-Term Memory. [Paper] [Code] AAAI 2024
  5. Agent Workflow Memory. [Paper] [Code] ICML 2025
  6. Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory. [Paper] [Code] ECAI 2025
  7. A-Mem: Agentic Memory for LLM Agents. [Paper] [Code] NeurIPS 2025
  8. HiAgent: Hierarchical Working Memory Management for Solving Long-Horizon Agent Tasks with Large Language Model. [Paper] ACL 2025
  9. Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models. [Paper] [Code] ICLR 2026
  10. Dynamic Affective Memory Management for Personalized LLM Agents. [Paper] arXiv 2025
  11. Meta Context Engineering via Agentic Skill Evolution. [Paper] [Code] ICML 2026
  12. AgentSys: Secure and Dynamic LLM Agents Through Explicit Hierarchical Memory Management. [Paper] arXiv 2026

3.2. Agentic Training

3.2.1. Environment Construction

Rule-Based Environments
  1. ALFWorld: Aligning Text and Embodied Environments for Interactive Learning. [Paper] [Code] ICLR 2021
  2. ScienceWorld: Is your Agent Smarter than a 5th Grader? [Paper] [Code] EMNLP 2022
  3. WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents. [Paper] [Code] NeurIPS 2022
  4. InterCode: Standardizing and Benchmarking Interactive Coding with Execution Feedback. [Paper] [Code] NeurIPS 2023
  5. AlphaMath Almost Zero: Process Supervision without Process. [Paper] NeurIPS 2024
  6. Step-level Value Preference Optimization for Mathematical Reasoning. [Paper] Findings of EMNLP 2024
  7. AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents. [Paper] [Code] ACL 2024
  8. ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery. [Paper] [Code] ICLR 2025
  9. MLGym: A New Framework and Benchmark for Advancing AI Research Agents. [Paper] arXiv 2025
  10. Reinforcement Learning for Reasoning in Large Language Models with One Training Example. [Paper] arXiv 2025
  11. FormalMATH: Benchmarking Formal Mathematical Reasoning of Large Language Models. [Paper] arXiv 2025
  12. REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewards. [Paper] arXiv 2025
  13. R-Zero: Self-Evolving Reasoning LLM from Zero Data. [Paper] arXiv 2025
  14. Immersion in the GitHub Universe: Scaling Coding Agents to Mastery. [Paper] arXiv 2026
  15. EnterpriseOps-Gym: Environments and Evaluations for Stateful Agentic Planning and Tool Use in Enterprise Settings. [Paper] arXiv 2026
Simulation-Based Environments
  1. Reasoning with Language Model is Planning with World Model. [Paper] EMNLP 2023
  2. NeuralOS: Towards Simulating Operating Systems via Neural Generative Models. [Paper] arXiv 2025
  3. BuilderBench: The Building Blocks of Intelligent Agents. [Paper] arXiv 2025
  4. MEAL: A Benchmark for Continual Multi-Agent Reinforcement Learning. [Paper] arXiv 2025
  5. Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning. [Paper] arXiv 2026
  6. WebWorld: A Large-Scale World Model for Web Agent Training. [Paper] arXiv 2026
Real-World Environments
  1. PaLM-E: An Embodied Multimodal Language Model. [Paper] ICML 2023
  2. RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control. [Paper] CoRL 2023
  3. WebArena: A Realistic Web Environment for Building Autonomous Agents. [Paper] [Code] ICLR 2024
  4. AgentBench: Evaluating LLMs as Agents. [Paper] [Code] ICLR 2024
  5. DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning. [Paper] [Code] NeurIPS 2024
  6. The BrowserGym Ecosystem for Web Agent Research. [Paper] TMLR 2025
  7. Digi-Q: Learning Q-Value Functions for Training Device-Control Agents. [Paper] ICLR 2025
  8. WebRL: Training LLM Web Agents via Self-Evolving Online Curriculum Reinforcement Learning. [Paper] [Code] ICLR 2025
  9. DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments. [Paper] [Code] EMNLP 2025
  10. PhysiAgent: An Embodied Agent Framework in Physical World. [Paper] arXiv 2025
  11. BrowseComp-V^3: A Visual, Vertical, and Verifiable Benchmark for Multimodal Browsing Agents. [Paper] arXiv 2026
  12. EnterpriseBench Corecraft: Training Generalizable Agents on High-Fidelity RL Environments. [Paper] arXiv 2026
  13. MolmoWeb: Open Visual Web Agent and Open Data for the Open Web. [Paper] arXiv 2026

3.2.2. Reward Design

Outcome-Level Rewards
  1. DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search. [Paper] ICLR 2025
  2. ACECODER: Acing Coder RL via Automated Test-Case Synthesis. [Paper] ACL 2025
  3. Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning. [Paper] [Code] COLM 2025
  4. TTRL: Test-Time Reinforcement Learning. [Paper] arXiv 2025
  5. Agentic Reasoning and Tool Integration for LLMs via Reinforcement Learning. [Paper] arXiv 2025
  6. Agent RL Scaling Law: Agent RL with Spontaneous Code Execution for Mathematical Problem Solving. [Paper] arXiv 2025
  7. Qwen3 Technical Report. [Paper] arXiv 2025
  8. Outcome-based Reinforcement Learning to Predict the Future. [Paper] TMLR 2025
  9. DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments. [Paper] [Code] EMNLP 2025
  10. DeepCoder: A Fully Open-Source 14B Coder at O3-mini Level. [Blog] Blog 2025
  11. DeepSWE: Training a Fully Open-sourced, State-of-the-Art Coding Agent by Scaling RL. [Blog] Blog 2025
Process-Level Rewards
  1. Let's Verify Step by Step. [Paper] [Code] ICLR 2024
  2. Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations. [Paper] ACL 2024
  3. Improve Mathematical Reasoning in Language Models by Automated Process Supervision. [Paper] arXiv 2024
  4. ToolRL: Reward is All Tool Learning Needs. [Paper] [Code] NeurIPS 2025
  5. GUI-G^2: Gaussian Reward Modeling for GUI Grounding. [Paper] AAAI 2026
  6. Process Reward Models That Think. [Paper] [Code] TMLR 2026

3.2.3. Training Optimization Algorithms

Supervised Fine-Tuning
  1. Toolformer: Language Models Can Teach Themselves to Use Tools. [Paper] NeurIPS 2023
  2. Gorilla: Large Language Model Connected with Massive APIs. [Paper] [Code] NeurIPS 2024
  3. ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs. [Paper] [Code] ICLR 2024
  4. FireAct: Toward Language Agent Fine-tuning. [Paper] [Code] arXiv 2023
  5. AgentTuning: Enabling Generalized Agent Abilities for LLMs. [Paper] [Code] Findings of ACL 2024
  6. AgentOhana: Design Unified Data and Training Pipeline for Effective Agent Learning. [Paper] [Code] arXiv 2024
  7. CodeAct: Executable Code Actions Elicit Better LLM Agents. [Paper] [Code] ICML 2024
  8. Agent-FLAN: Designing Data and Methods of Effective Agent Tuning for Large Language Models. [Paper] [Code] Findings of ACL 2024
  9. MAGDi: Structured Distillation of Multi-Agent Interaction Graphs Improves Reasoning in Smaller Language Models. [Paper] ICML 2024
  10. On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes. [Paper] ICLR 2024
  11. MiniLLM: On-Policy Distillation of Large Language Models. [Paper] ICLR 2024
  12. AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials. [Paper] [Code] ICLR 2025
  13. Efficient Agent Training for Computer Use. [Paper] arXiv 2025
  14. Structured Agent Distillation for Large Language Model. [Paper] arXiv 2025
  15. Merge-of-Thought Distillation. [Paper] arXiv 2025
  16. Black-Box On-Policy Distillation of Large Language Models. [Paper] arXiv 2025
  17. Expanding the Capability Frontier of LLM Agents with ZPD-Guided Data Synthesis. [Paper] ICLR 2026
  18. Stable On-Policy Distillation through Adaptive Target Reformulation. [Paper] arXiv 2026
  19. Learning beyond Teacher: Generalized On-Policy Distillation with Reward Extrapolation. [Paper] arXiv 2026
  20. On Data Engineering for Scaling LLM Terminal Capabilities. [Paper] arXiv 2026
  21. Scaling Reasoning Efficiently via Relaxed On-Policy Distillation. [Paper] arXiv 2026
  22. Immersion in the GitHub Universe: Scaling Coding Agents to Mastery. [Paper] arXiv 2026
  23. ClawGym: A Scalable Framework for Building Effective Claw Agents. [Paper] arXiv 2026
Reinforcement Learning Approaches
  1. Proximal Policy Optimization Algorithms. [Paper] arXiv 2017
  2. Direct Preference Optimization: Your Language Model is Secretly a Reward Model. [Paper] [Code] NeurIPS 2023
  3. Secrets of RLHF in Large Language Models Part I: PPO. [Paper] [Code] arXiv 2023
  4. LATS: Language Agent Tree Search Unifies Reasoning, Acting, and Planning. [Paper] [Code] ICML 2024
  5. Secrets of RLHF in Large Language Models Part II: Reward Modeling. [Paper] [Code] arXiv 2024
  6. KTO: Model Alignment as Prospect Theoretic Optimization. [Paper] [Code] arXiv 2024
  7. ArCHer: Training Language Model Agents via Hierarchical Multi-Turn RL. [Paper] [Code] ICML 2024
  8. DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models. [Paper] [Code] arXiv 2024
  9. ORPO: Monolithic Preference Optimization without Reference Model. [Paper] [Code] EMNLP 2024
  10. SimPO: Simple Preference Optimization with a Reference-Free Reward. [Paper] [Code] NeurIPS 2024
  11. WARP: On the Benefits of Weight Averaged Rewarded Policies. [Paper] arXiv 2024
  12. ReMax: A Simple, Effective, and Efficient Reinforcement Learning Method for Aligning Large Language Models. [Paper] ICML 2024
  13. VinePPO: Refining Credit Assignment in RL Training of LLMs. [Paper] ICML 2025
  14. Kimi k1.5: Scaling Reinforcement Learning with LLMs. [Paper] arXiv 2025
  15. DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning. [Paper] Nature 2025
  16. DAPO: An Open-Source LLM Reinforcement Learning System at Scale. [Paper] [Code] arXiv 2025
  17. RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning. [Paper] [Code] arXiv 2025
  18. WebThinker: Empowering Large Reasoning Models with Deep Research Capability. [Paper] [Code] NeurIPS 2025
  19. Group-in-Group Policy Optimization for LLM Agent Training. [Paper] [Code] arXiv 2025
  20. Agentic Reinforced Policy Optimization. [Paper] [Code] arXiv 2025
  21. LLM Collaboration with Multi-Agent Reinforcement Learning. [Paper] AAAI 2026
  22. Tree Search for LLM Agent Reinforcement Learning. [Paper] arXiv 2025
  23. ML-Agent: Reinforcing LLM Agents for Autonomous Machine Learning Engineering. [Paper] arXiv 2025
  24. ACT: Agentic Critical Training. [Paper] arXiv 2026

3.2.4. Infrastructure

General-Purpose Frameworks
  1. OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework. [Paper] [Code] arXiv 2024
  2. NeMo-Aligner: Scalable Toolkit for Efficient Model Alignment. [Paper] [Code] arXiv 2024
  3. ReaL: Efficient RLHF Training of Large Language Models with Parameter Reallocation. [Paper] [Code] arXiv 2024
  4. AgentGym: Evolving Large Language Model-based Agents across Diverse Environments. [Paper] [Code] arXiv 2024
  5. HybridFlow: A Flexible and Efficient RLHF Framework. [Paper] [Code] EuroSys 2025
  6. Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Model. [Paper] arXiv 2025
  7. RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning. [Paper] [Code] arXiv 2025
  8. Group-in-Group Policy Optimization for LLM Agent Training. [Paper] [Code] arXiv 2025
  9. AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning. [Paper] [Code] arXiv 2025
  10. Agent Lightning: Train ANY AI Agents with Reinforcement Learning. [Paper] [Code] arXiv 2025
  11. AgentGym-RL: Training LLM Agents for Long-Horizon Decision Making through Multi-Turn Reinforcement Learning. [Paper] arXiv 2025
  12. GEM: A Gym for Agentic LLMs. [Paper] [Code] arXiv 2025
  13. AgentRL: Scaling Agentic Reinforcement Learning with a Multi-Turn, Multi-Task Framework. [Paper] arXiv 2025
  14. SkyRL-Agent: Efficient RL Training for Multi-turn LLM Agent. [Paper] [Code] arXiv 2025
  15. Agent-R1: Training Powerful LLM Agents with End-to-End Reinforcement Learning. [Paper] [Code] arXiv 2025
  16. RollArt: Scaling Agentic RL Training via Disaggregated Infrastructure. [Paper] arXiv 2025
Specialized Frameworks
  1. Multimodal Reinforcement Learning with Agentic Verifier for AI Agents. [Paper] arXiv 2025
  2. Dr. MAS: Stable Reinforcement Learning for Multi-Agent LLM Systems. [Paper] arXiv 2026
  3. MARTI: A Framework for Multi-Agent LLM Systems Reinforced Training and Inference. [Paper] ICLR 2026
  4. LLM Collaboration with Multi-Agent Reinforcement Learning. [Paper] AAAI 2026
  5. WideSeek-R1: Exploring Width Scaling for Broad Information Seeking via Multi-Agent Reinforcement Learning. [Paper] arXiv 2026
  6. MM-DeepResearch: A Simple and Effective Multimodal Agentic Search Baseline. [Paper] arXiv 2026
  7. GrandCode: Achieving Grandmaster Level in Competitive Programming via Agentic Reinforcement Learning. [Paper] arXiv 2026
  8. LiteResearcher: A Scalable Agentic RL Training Framework for Deep Research Agent. [Paper] [Code] arXiv 2026

4. Representative Benchmarks by Task Domain

4.1. Deep Research

  1. BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agents. [Paper] [Code] arXiv 2025
  2. IDRBench: Interactive Deep Research Benchmark. [Paper] arXiv 2026
  3. ReportBench: Evaluating Deep Research Agents via Academic Survey Tasks. [Paper] [Code] arXiv 2025
  4. Characterizing Deep Research: A Benchmark and Formal Definition. [Paper] [Code] ICLR 2026
  5. WideSearch: Benchmarking Agentic Broad Info-Seeking. [Paper] [Code] ICLR 2026
  6. LiveResearchBench: A Live Benchmark for User-Centric Deep Research in the Wild. [Paper] [Code] ICLR 2026
  7. ResearchRubrics: A Benchmark of Prompts and Rubrics For Evaluating Deep Research Agents. [Paper] [Code] ICLR 2026
  8. DeepSynth-Eval: Objectively Evaluating Information Consolidation in Deep Survey Writing. [Paper] arXiv 2026
  9. DeepResearch Bench II: Diagnosing Deep Research Agents via Rubrics from Expert Report. [Paper] [Code] arXiv 2026
  10. MMDeepResearch-Bench: A Benchmark for Multimodal Deep Research Agents. [Paper] [Code] arXiv 2026
  11. DeepSurvey-Bench: Evaluating Academic Value of Automatically Generated Scientific Survey. [Paper] arXiv 2026

4.2. Software Engineering

  1. RepoBench: Benchmarking Repository-Level Code Auto-Completion Systems. [Paper] [Code] ICLR 2024
  2. SWE-bench: Can Language Models Resolve Real-world Github Issues? [Paper] [Code] ICLR 2024
  3. LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code. [Paper] [Code] ICLR 2025
  4. Can Language Models Replace Programmers for Coding? REPOCOD Says 'Not Yet'. [Paper] [Code] ACL 2025
  5. SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks? [Paper] [Code] arXiv 2025
  6. SWE-Sharp-Bench: A Reproducible Benchmark for C# Software Engineering Tasks. [Paper] [Code] AIware 2025
  7. LoCoBench-Agent: An Interactive Benchmark for LLM Agents in Long-Context Software Engineering. [Paper] [Code] arXiv 2025
  8. NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents. [Paper] [Code] ICML 2026
  9. OmniCode: A Benchmark for Evaluating Software Engineering Agents. [Paper] [Code] arXiv 2026
  10. SWE-Universe: Scale Real-World Verifiable Environments to Millions. [Paper] arXiv 2026
  11. FeatureBench: Benchmarking Agentic Coding for Complex Feature Development. [Paper] [Code] ICLR 2026

4.3. Tool Use and Function Calling

  1. API-Bank: A Comprehensive Benchmark for Tool-Augmented LLMs. [Paper] [Code] EMNLP 2023
  2. AgentBench: Evaluating LLMs as Agents. [Paper] [Code] ICLR 2024
  3. GAIA: a benchmark for General AI Assistants. [Paper] [Code] ICLR 2024
  4. tau-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains. [Paper] [Code] ICLR 2025
  5. AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents. [Paper] [Code] ACL 2024
  6. AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks? [Paper] [Code] EMNLP 2024
  7. Toolsandbox: A stateful, conversational, interactive evaluation benchmark for llm tool use capabilities. [Paper] [Code] Findings of NAACL 2025
  8. ToolHop: A Query-Driven Benchmark for Evaluating Large Language Models in Multi-Hop Tool Use. [Paper] [Code] ACL 2025
  9. DICE-BENCH: Evaluating the Tool-Use Capabilities of Large Language Models in Multi-Round, Multi-Party Dialogues. [Paper] [Code] Findings of ACL 2025
  10. tau^2-Bench: Evaluating Conversational Agents in a Dual-Control Environment. [Paper] [Code] arXiv 2025
  11. The Berkeley Function Calling Leaderboard (BFCL): From Tool Use to Agentic Evaluation of Large Language Models. [Paper] [Code] ICML 2025
  12. CCTU: A Benchmark for Tool Use under Complex Constraints. [Paper] [Code] arXiv 2026

4.4. Computer Use and GUI Grounding

  1. Mind2Web: Towards a Generalist Agent for the Web. [Paper] [Code] NeurIPS 2023
  2. WebArena: A Realistic Web Environment for Building Autonomous Agents. [Paper] [Code] ICLR 2024
  3. Android in the Wild: A Large-Scale Dataset for Android Device Control. [Paper] [Code] NeurIPS 2023
  4. VisualWebArena: Evaluating Multimodal Agents on Realistic Visual Web Tasks. [Paper] [Code] ACL 2024
  5. WorkArena: How Capable are Web Agents at Solving Common Knowledge Work Tasks? [Paper] [Code] ICML 2024
  6. AndroidWorld: A Dynamic Benchmarking Environment for Autonomous Agents. [Paper] [Code] ICLR 2025
  7. Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale. [Paper] [Code] ICML 2025
  8. WorldGUI: An Interactive Benchmark for Desktop GUI Automation from Any Starting Point. [Paper] [Code] arXiv 2025
  9. ScreenSpot-Pro: GUI Grounding for Professional High-Resolution Computer Use. [Paper] [Code] ICLR 2025 Workshop
  10. MMBench-GUI: Hierarchical Multi-Platform Evaluation Framework for GUI Agents. [Paper] [Code] arXiv 2025
  11. Osworld-mcp: Benchmarking mcp tool invocation in computer-use agents. [Paper] [Code] ICLR 2026
  12. VenusBench-GD: A Comprehensive Multi-Platform GUI Benchmark for Diverse Grounding Tasks. [Paper] [Code] arXiv 2025

4.5. ML Engineering and Scientific Research

  1. DS-1000: A Natural and Reliable Benchmark for Data Science Code Generation. [Paper] [Code] ICML 2023
  2. MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation. [Paper] [Code] ICML 2024
  3. DA-Code: Agent Data Science Code Generation Benchmark for Large Language Models. [Paper] [Code] EMNLP 2024
  4. ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discovery. [Paper] [Code] ICLR 2025
  5. MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering. [Paper] [Code] ICLR 2025
  6. PaperBench: Evaluating AI's Ability to Replicate AI Research. [Paper] [Code] ICML 2025
  7. TimeSeriesGym: A Scalable Benchmark for (Time Series) Machine Learning Engineering Agents. [Paper] [Code] NeurIPS 2025 Workshop
  8. MLR-Bench: Evaluating AI Agents on Open-Ended Machine Learning Research. [Paper] [Code] NeurIPS 2025
  9. FML-bench: Benchmarking Machine Learning Agents for Scientific Research. [Paper] [Code] arXiv 2025
  10. ReplicatorBench: Benchmarking LLM Agents for Replicability in Social and Behavioral Sciences. [Paper] [Code] MSLD 2026

5. Future Directions

5.1. Efficiency

  1. RouteLLM: Learning to Route LLMs from Preference Data. [Paper] ICLR 2025
  2. Chain of Thoughtlessness? An Analysis of CoT in Planning. [Paper] [Code] NeurIPS 2024
  3. ACON: Optimizing Context Compression for Long-horizon LLM Agents. [Paper] [Code] arXiv 2025
  4. Reducing Cost of LLM Agents with Trajectory Reduction. [Paper] arXiv 2025
  5. ARC: Active and Reflection-driven Context Management for Long-Horizon Information Seeking Agents. [Paper] arXiv 2026
  6. Context as a Tool: Context Management for Long-Horizon SWE-Agents. [Paper] arXiv 2025
  7. Robust and Efficient Tool Orchestration via Layered Execution Structures with Reflective Correction. [Paper] arXiv 2026

5.2. Safety

  1. AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents. [Paper] NeurIPS 2024
  2. MCPTox: A Benchmark for Tool Poisoning Attack on Real-World MCP Servers. [Paper] arXiv 2025
  3. AgentSys: Secure and Dynamic LLM Agents Through Explicit Hierarchical Memory Management. [Paper] arXiv 2026
  4. OpenPort Protocol: A Security Governance Specification for AI Agent Tool Access. [Paper] arXiv 2026
  5. Runtime Governance for AI Agents: Policies on Paths. [Paper] arXiv 2026
  6. AgentSpec: Customizable runtime enforcement for safe and reliable LLM agents.(2026). [Paper] ICSE 2026
  7. Autonomous Action Runtime Management(AARM):A System Specification for Securing AI-Driven Actions at Runtime. [Paper] arXiv 2026
  8. SafeHarness: Lifecycle-Integrated Security Architecture for LLM-based Agent Deployment. [Paper] arXiv 2026

5.3. Continual Learning

  1. MemGPT: Towards LLMs as Operating Systems. [Paper] [Code] arXiv 2023
  2. A-Mem: Agentic Memory for LLM Agents. [Paper] [Code] NeurIPS 2025
  3. Memory as Action: Autonomous Context Curation for Long-Horizon Agentic Tasks. [Paper] arXiv 2025
  4. Context Engineering: Memory, Compaction, and Tool Clearing. [Docs] Docs 2026
  5. Agent Workflow Memory. [Paper] [Code] ICML 2025

5.4. State and Environment Modeling

  1. WorldCoder, a Model-Based LLM Agent: Building World Models by Writing Code and Interacting with the Environment. [Paper] NeurIPS 2024
  2. PABU: Progress-Aware Belief Update for Efficient LLM Agents. [Paper] arXiv 2026
  3. Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence. [Paper] arXiv 2026

5.5. Embodied Harnesses

  1. PaLM-E: An Embodied Multimodal Language Model. [Paper] ICML 2023
  2. RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control. [Paper] CoRL 2023
  3. Green-VLA: Staged Vision-Language-Action Model for Generalist Robots. [Paper] arXiv 2026
  4. ACoT-VLA: Action Chain-of-Thought for Vision-Language-Action Models. [Paper] arXiv 2026
  5. ST4VLA: Spatially Guided Training for Vision-Language-Action Models. [Paper] arXiv 2026
  6. RoboAgent: Chaining Basic Capabilities for Embodied Task Planning. [Paper] CVPR 2026
  7. EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents. [Paper] ICML 2025
  8. AgentSpec: Customizable Runtime Enforcement for Safe and Reliable LLM Agents. [Paper] ICSE 2026

5.6. Evaluation

  1. BrowseComp-V^3: A Visual, Vertical, and Verifiable Benchmark for Multimodal Browsing Agents. [Paper] arXiv 2026
  2. OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments. [Paper] [Code] NeurIPS 2024
  3. PaperBench: Evaluating AI's Ability to Replicate AI Research. [Paper] [Code] ICML 2025
  4. ReplicatorBench: Benchmarking LLM Agents for Replicability in Social and Behavioral Sciences. [Paper] [Code] MSLD 2026
  5. BeyondSWE: Can Current Code Agent Survive Beyond Single-Repo Bug Fixing? [Paper] arXiv 2026
  6. Gaia-2: A controllable multi-view generative world model for autonomous driving. [Paper] arXiv 2025
  7. Claw-Eval-Live: A Live Agent Benchmark for Evolving Real-World Workflows. [Paper] arXiv 2026
  8. ResearchRubrics: A Benchmark of Prompts and Rubrics For Evaluating Deep Research Agents. [Paper] [Code] ICLR 2026
  9. Osworld-mcp: Benchmarking mcp tool invocation in computer-use agents. [Paper] [Code] ICLR 2026
  10. AgentLongBench: A Controllable Long Benchmark For Long-Contexts Agents via Environment Rollouts. [Paper] arXiv 2026
  11. ATBench: A Diverse and Realistic Agent Trajectory Benchmark for Safety Evaluation and Diagnosis. [Paper] arXiv 2026
  12. AgentLeak: A Full-Stack Benchmark for Privacy Leakage in Multi-Agent LLM Systems. [Paper] arXiv 2026
  13. Benchmark Test-Time Scaling of General LLM Agents. [Paper] arXiv 2026

Citation

If you find our paper useful for your research, please consider citing the paper and starring this repository.

@misc{tang2026agent,
  title={Agent Systems with Harness Engineering},
  author={Xinyu Tang and Han Peng and Guoxin Chen and Yuze Shi and
          Zitao Su and Peiyu Liu and Wayne Xin Zhao and Yawen Li and
          Zhe Xue},
  url={https://openreview.net/pdf?id=nM5tDHrQsx},
  year={2026}
}