Toward Training Superintelligent Software Agents through Self-Play SWE-RL

• URL: http://arxiv.org/abs/2512.18552v1
• Date: Sun, 21 Dec 2025 00:49:40 GMT
• 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,
• Abstract summary: Self-play SWE-RL is a first step toward training paradigms for superintelligent software agents.<n>Our approach takes minimal data assumptions, only requiring access to sandboxed repositories with source code and installed dependencies.<n>Our results, albeit early, suggest a path where agents autonomously gather extensive learning experiences from real-world software repositories.
• Score: 66.11447353341926
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• AI IDEs or Autonomous Agents? Measuring the Impact of Coding Agents on Software Development [12.50615284537175]
  Large language model (LLM) based coding agents increasingly act as autonomous contributors that generate and merge pull requests.<n>We present a longitudinal causal study of agent adoption in open-source repositories using staggered difference-in-differences with matched controls.
  arXiv  Detail & Related papers  (2026-01-20T04:51:56Z)
• Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning [84.70211451226835]
  Large Language Model (LLM) Agents are constrained by a dependency on human-curated data.<n>We introduce Agent0, a fully autonomous framework that evolves high-performing agents without external data.<n>Agent0 substantially boosts reasoning capabilities, improving the Qwen3-8B-Base model by 18% on mathematical reasoning and 24% on general reasoning benchmarks.
  arXiv  Detail & Related papers  (2025-11-20T05:01:57Z)
• SFR-DeepResearch: Towards Effective Reinforcement Learning for Autonomously Reasoning Single Agents [93.26456498576181]
  This paper focuses on the development of native Autonomous Single-Agent models for Deep Research.<n>Our best variant SFR-DR-20B achieves up to 28.7% on Humanity's Last Exam benchmark.
  arXiv  Detail & Related papers  (2025-09-08T02:07:09Z)
• Re:Form -- Reducing Human Priors in Scalable Formal Software Verification with RL in LLMs: A Preliminary Study on Dafny [78.1575956773948]
  Large Language Models (LLMs) trained with Reinforcement Learning (RL) face a significant challenge: their verification processes are neither reliable nor scalable.<n>A promising yet largely uncharted alternative is formal language-based reasoning.<n>Grounding LLMs in rigorous formal systems where generative models operate in formal language spaces (e.g., Dafny) enables the automatic and mathematically provable verification of their reasoning processes and outcomes.
  arXiv  Detail & Related papers  (2025-07-22T08:13:01Z)
• SWE-Bench-CL: Continual Learning for Coding Agents [0.0]
  SWE-Bench-CL is a novel continual learning benchmark built on the human-verified SWE-Bench Verified dataset.<n>By organizing GitHub issues into chronologically ordered sequences that reflect natural repository evolution, SWE-Bench-CL enables direct evaluation of an agent's ability to accumulate experience.
  arXiv  Detail & Related papers  (2025-06-13T07:11:14Z)
• Exploring RL-based LLM Training for Formal Language Tasks with Programmed Rewards [49.7719149179179]
  This paper investigates the feasibility of using PPO for reinforcement learning (RL) from explicitly programmed reward signals. We focus on tasks expressed through formal languages, such as programming, where explicit reward functions can be programmed to automatically assess quality of generated outputs. Our results show that pure RL-based training for the two formal language tasks is challenging, with success being limited even for the simple arithmetic task.
  arXiv  Detail & Related papers  (2024-10-22T15:59:58Z)
• Evaluating Software Development Agents: Patch Patterns, Code Quality, and Issue Complexity in Real-World GitHub Scenarios [13.949319911378826]
  This study evaluated 4,892 patches from 10 top-ranked agents on 500 real-world GitHub issues.<n>No single agent dominated, with 170 issues unresolved, indicating room for improvement.<n>Most agents maintained code reliability and security, avoiding new bugs or vulnerabilities.<n>Some agents increased code complexity, many reduced code duplication and minimized code smells.
  arXiv  Detail & Related papers  (2024-10-16T11:33:57Z)
• Multi-Agent Reinforcement Learning for Autonomous Driving: A Survey [14.73689900685646]
  Reinforcement Learning (RL) is a potent tool for sequential decision-making and has achieved performance surpassing human capabilities. As the extension of RL in the multi-agent system domain, multi-agent RL (MARL) not only need to learn the control policy but also requires consideration regarding interactions with all other agents in the environment. Simulators are crucial to obtain realistic data, which is the fundamentals of RL.
  arXiv  Detail & Related papers  (2024-08-19T03:31:20Z)
• Agentless: Demystifying LLM-based Software Engineering Agents [12.19683999553113]
  We build Agentless -- an agentless approach to automatically solve software development problems. Compared to the verbose and complex setup of agent-based approaches, Agentless employs a simplistic three-phase process of localization, repair, and patch validation. Our results on the popular SWE-bench Lite benchmark show that surprisingly the simplistic Agentless is able to achieve both the highest performance and low cost.
  arXiv  Detail & Related papers  (2024-07-01T17:24:45Z)
• Symbolic Learning Enables Self-Evolving Agents [55.625275970720374]
  We introduce agent symbolic learning, a systematic framework that enables language agents to optimize themselves on their own. Agent symbolic learning is designed to optimize the symbolic network within language agents by mimicking two fundamental algorithms in connectionist learning. We conduct proof-of-concept experiments on both standard benchmarks and complex real-world tasks.
  arXiv  Detail & Related papers  (2024-06-26T17:59:18Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.