[2512.18552] Toward Training Superintelligent Software Agents through Self-Play SWE-RL

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs. arXiv Operational Status

[2512.18552] Toward Training Superintelligent Software Agents through Self-Play SWE-RL --> Computer Science > Software Engineering arXiv:2512.18552 (cs) [Submitted on 21 Dec 2025 ( v1 ), last revised 2 Jun 2026 (this version, v3)] Title: Toward Training Superintelligent Software Agents through Self-Play SWE-RL Authors: Yuxiang Wei , Zhiqing Sun , Emily McMilin , Jonas Gehring , David Zhang , Gabriel Synnaeve , Daniel Fried , Lingming Zhang , Sida Wang View a PDF of the paper titled Toward Training Superintelligent Software Agents through Self-Play SWE-RL, by Yuxiang Wei and 8 other authors Vie

saved by

• 4678

related reading

• 2502.18449arxiv.org
• Composer2.pdfcursor.com
• The Era of Experience Paper.pdfstorage.googleapis.com
• LLM Powered Autonomous Agents | Lil'Loglilianweng.github.io
• Automated Weak-to-Strong Researcheralignment.anthropic.com
• [2603.08640] PostTrainBench: Can LLM Agents Automate LLM Post-Training?arxiv.org
• The Eternal Sloptember | the singularity is nearergeohot.github.io
• pdfopenreview.net
• Introducing SWE-grep and SWE-grep-mini: RL for Multi-Turn, Fast Context Retrieval | Cognitioncognition.ai
• Andrej Karpathy — AGI is still a decade awaydwarkesh.com
• Towards self-driving codebases · Cursorcursor.com
• [2604.22082] Removing Sandbagging in LLMs by Training with Weak Supervisionarxiv.org