🕹 Benchmarks

CLI

• [Preprint'26 - framework/terminal-bench)
• [Website - framework/terminal-bench-science)

Computer Use

• [Paper - Pro-GUI-Grounding)
• [NeurIPS'24 - ai/OSWorld)
• [ACL'24
• [Platform - ai/computer-agent-arena)
• [Preprint'26 - AI-Lab/ClawBench) [Website]
• [Preprint'26 - ai/OSWorld-V2)
• [Preprint'26 - berkeley/agents-last-exam) [Website]

Deep research

• [Blog - d6f7-442e-9508-59515c65e35d/browsecomp.pdf) [Code]
• [Preprint'25 - ZH)
• [Preprint'25
• [Preprint'25 - plus)
• [Preprint'25
• [ICLR'26 - BrowseComp)
• [Preprint'25(1) - AI/Marco-Search-Agent)
• [Preprint'26 - DeepResearch)
• [Preprint'26 - nlp/AgentVista)
• [Preprint'26 - Benchmark)
• [Preprint'25 - plus/MMSearch-Plus)
• [Preprint'26 - NLPIR/OmniGAIA)

🧪 Demos & Projects

MCP Agents

• [Code
• [Code

RL-based LLM agent tuning

• [Code
• [Blog - AI/SkyRL)
• [Code
• [Code
• [Code
• [Page - R1)
• [Code
• [Code

RL-based LLM tuning

• [Preprint'25 - Reasoner-Zero/Open-Reasoner-Zero)
• [Code

📚 Papers & Research

Others

• [Preprint'25
• [Preprint'25
• [Preprint'25 - feedback)
• [Code
• [Blog - dots-ai/dots3-note-prev)
• [Preprint'26 - uofa/Self-Distillation-AgentRL)

Reinforcement Learning Scaling

• [Preprint'25
• [Preprint'25 - SIA/DAPO)
• [Preprint'25 - NLP/LIMR)
• [Preprint'25
• [Preprint'25
• [Blog - OR1)
• [Preprint'25
• [Blog
• [Preprint'25 - R1V2-38B)
• [Preprint'25
• [Preprint'25

RL for Agent Memory

• [Preprint'25
• [Preprint'25 - SIA/MemAgent)

RL for Computer-using Agents

• [Preprint'25 - R1)
• [Preprint'25 - agent/digiq)
• [Preprint'25 - Labs/InfiGUI-R1)
• [Preprint'25
• [KDD'24
• [Preprint'25 - research/ARPO)
• [Preprint'25 - ai/OpenCUA)
• [Preprint'25 - TARS)
• [Preprint'26 - org.com/blog/tmax/#training-tmax-with-rl)

RL for Multi-Modal Agent (Thinking w Image / MMSearch)

• [Preprint'25 - Lab/multimodal-search-r1)
• [ICLR'26 - Agent/DeepEyesV2)
• [ICLR'26 - Agent/DeepEyes)
• [Preprint'26 - without-Zooming)

RL for Research Agents

• [Preprint'25 - Reasoning)
• [Preprint'25 - nlp/ZeroSearch)
• [EMNLP'25 - NLP/DeepResearcher)
• [Preprint'25 - RL/ReCall)
• [COLM'25 - r1)
• [Preprint'25 - NLP/WebAgent)
• [Blog
• [Preprint'25 - Search)
• [Preprint'25 - Searcher-plus)
• [Preprint'25 - Searcher)
• [Preprint'25 - NLP/WebAgent)
• [Blog
• [Preprint'25 - Pro)
• [Preprint'25
• [Preprint'25
• [Preprint'25
• [Blog - NLP/DeepResearch)
• [Preprint'25
• [Preprint'25
• [Preprint'26
• [Preprint'26

RL for Tool-using Problem Solver

• [Preprint'25 - BJTU/AutoCoA)
• [Preprint'25 - NLP/ToRL)
• [Preprint'25
• [Preprint'25
• [Preprint'25
• [Preprint'25
• [Preprint'25 - AI-Lab/verl-tool)
• [Preprint'25
• [Preprint'26

RL with Agent Skills

• [Preprint'26 - lab/SkillRL)
• [Preprint'26 - Aware-Training/) — Reinforcement learning across augmented skills, tools, prompts, and hooks for adaptation to harness changes.

Self-Playing Agent with RL

• [ICLR'26 - Quark/SSP)
• [Preprint'25 - lab/Agent0)
• [Preprint'25 - lab/Agent0)
• [Preprint'25

Survey & Review

• [Preprint'25 - agents-2030/awesome-deep-research-agent)
• [Preprint'25 - Search-Agent-Papers)
• [Preprint'25 - AgenticLLM-RL-Papers)

🔗 Related Awesome Lists

MCP Agents

• [List - covering RL for research agents
• [List - covering rl and agents before 2023
• [List - covering deep research agents and benchmark results
• [List - covering search agent papers
• [List - covering Agentic RL papers in both agentic capabilities and applications
• [List - covering agent memory papers
• [List - covering Agentic RL papers in agentic search systems

🧰 Toolkits & Frameworks

MCP Agents

• [Code
• [Code
• [Code
• [Code
• [Code
• [Code
• [Website - credit-bench)

📄 Tutorials & Blog Posts

MCP Agents

• [Blog
• [Github
• [Blog
• [Blog
• [Blog