While current software agents powered by large language models (LLMs) and agentic reinforcement learning (RL) can boost programmer productivity, their training data (e.g., GitHub issues and pull requests) and environments (e.g., pass-to-pass and fail-to-pass tests) heavily depend on human knowledge or curation, posing a fundamental barrier to superintelligence. In this paper, we present Self-play SWE-RL (SSR), a first step toward training paradigms for superintelligent software agents. Our approach takes minimal data assumptions, only requiring access to sandboxed repositories with source code and installed dependencies, with no need for human-labeled issues or tests. Grounded in these real-world codebases, a single LLM agent is trained via reinforcement learning in a self-play setting to iteratively inject and repair software bugs of increasing complexity, with each bug formally specified by a test patch rather than a natural language issue description. On the SWE-bench Verified and SWE-Bench Pro benchmarks, SSR achieves notable self-improvement (+10.4 and +7.8 points, respectively) and consistently outperforms the human-data baseline over the entire training trajectory, despite being evaluated on natural language issues absent from self-play. Our results, albeit early, suggest a path where agents autonomously gather extensive learning experiences from real-world software repositories, ultimately enabling superintelligent systems that exceed human capabilities in understanding how systems are constructed, solving novel challenges, and autonomously creating new software from scratch.
SWE-Dev: Evaluating and Training Autonomous Feature-Driven Software Development
Shanghai Jiao Tong University(上海交通大学);Beijing University of Aeronautics and Astronautics(北京航空航天大学);The Chinese University of Hong Kong(香港中文大学);University of Michigan(密歇根大学)
SWE-Search: Enhancing Software Agents with Monte Carlo Tree Search and Iterative Refinement
University of California, Santa Barbara(加利福尼亚大学圣巴巴拉分校);Moatless AI;Carnegie Mellon University(卡内基梅隆大学);National University of Singapore(新加坡国立大学);Mila(米拉研究所)
SWE-Marathon: Can Agents Autonomously Complete Ultra-Long-Horizon Software Work?
Abundant;Zenity;Harvard University(哈佛大学);University of Waterloo(滑铁卢大学);Gujarat Technological University(古吉拉特技术大学);Warping;Stanford University(斯坦福大学);UNC-Chapel Hill(北卡罗来纳大学教堂山分校);Independent(独立);Refresh;Soleda AI;Near AI;Georgia Tech(佐治亚理工学院);Comenius University in Bratislava(布拉迪斯拉发Comenius大学);UC San Diego(圣地亚哥大学);BenchFlow;UC Santa Barbara(圣巴巴拉大学)
Satori-SWE: Evolutionary Test-Time Scaling for Sample-Efficient Software Engineering
Singapore University of Technology and Design(新加坡科技设计大学);MIT(麻省理工学院);UMass Amherst(马萨诸塞大学阿默斯特分校);Harvard(哈佛大学);MIT-IBM Watson AI Lab(MIT-IBM沃森人工智能实验室);IBM Research(IBM研究院)
SWE-Chain: Benchmarking Coding Agents on Chained Release-Level Package Upgrades
The Chinese University of Hong Kong(香港中文大学);Independent(独立);ELLIS;Technical University of Darmstadt(达姆施塔特技术大学);Johns Hopkins University(约翰霍普金斯大学);Monash University(墨尔本大学)
SWE-Mutation: Can LLMs Generate Reliable Test Suites in Software Engineering?
State Key Laboratory of Cognitive Intelligence(认知智能国家重点实验室);University of Science and Technology of China(中国科学技术大学);Beihang University(北航);School of Mathematical Sciences, Peking University(北京大学数学科学学院);NeoShell AI;Institute of Artificial Intelligence, Hefei Comprehensive National Science Center(合肥综合性国家科学中心人工智能研究院)
Advances and Frontiers of LLM-based Issue Resolution in Software Engineering: A Comprehensive Survey
Sun Yat-sen University(中山大学);Hangzhou Normal University(杭州师范大学);Zhejiang University(浙江大学);Huawei Technologies Co, Ltd(华为技术有限公司);Chongqing University(重庆大学)
SWE-Edit: Rethinking Code Editing for Efficient SWE-Agent
Microsoft(微软);Department of Computer Sciences, University of Wisconsin–Madison(威斯康星大学麦迪逊分校计算机科学系);Stanford Institute for Human-Centered Artificial Intelligence (HAI), Stanford University(斯坦福大学人机交互研究所)