mlsanityData to 7 Oct 2026

SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution

arXiv 2502.18449 ↗First posted 25 Feb 2025cs.SE cs.AI cs.CLcited by 246

On the map: LLM reasoning → Reinforcement learning for reasoning

Abstract (arXiv)

The recent DeepSeek-R1 release has demonstrated the immense potential of reinforcement learning (RL) in enhancing the general reasoning capabilities of large language models (LLMs). While DeepSeek-R1 and other follow-up work primarily focus on applying RL to competitive coding and math problems, this paper introduces SWE-RL, the first approach to scale RL-based LLM reasoning for real-world software engineering. Leveraging a lightweight rule-based reward (e.g., the similarity score between ground-truth and LLM-generated solutions), SWE-RL enables LLMs to autonomously recover a developer's reasoning processes and solutions by learning from extensive open-source software evolution data -- the record of a software's entire lifecycle, including its code snapshots, code changes, and events such as issues and pull requests. Trained on top of Llama 3, our resulting reasoning model, Llama3-SWE-RL-70B, achieves a 41.0% solve rate on SWE-bench Verified -- a human-verified collection of real-world GitHub issues. To our knowledge, this is the best performance reported for medium-sized (<100B) LLMs to date, even comparable to leading proprietary LLMs like GPT-4o. Surprisingly, despite performing RL solely on software evolution data, Llama3-SWE-RL has even emerged with generalized reasoning skills. For example, it shows improved results on five out-of-domain tasks, namely, function coding, library use, code reasoning, mathematics, and general language understanding, whereas a supervised-finetuning baseline even leads to performance degradation on average. Overall, SWE-RL opens up a new direction to improve the reasoning capabilities of LLMs through reinforcement learning on massive software engineering data.

Key facts. Cited by 246 papers in the map (data through 7 Oct 2026). Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning. Builds on 33 of its 65 references that are in the map. First posted on arXiv 25 Feb 2025.

Builds on

33 of 65 references.

References · 33

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  3. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  5. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  6. 2021Program Synthesis with Large Language Modelscited by 3,509
  7. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 1,906
  8. 2024SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineeringcited by 1,628
  9. 2022Competition-Level Code Generation with AlphaCodecited by 1,427
  10. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,332
  11. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,154
  12. 2021Measuring Coding Challenge Competence With APPScited by 879
21 more
  1. 2024MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategiescited by 694
  2. 2024BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructionscited by 527
  3. 2022CodeT: Code Generation with Generated Testscited by 369
  4. 2023RepoBench: Benchmarking Repository-Level Code Auto-Completion Systemscited by 346
  5. 2024Training Software Engineering Agents and Verifiers with SWE-Gymcited by 300
  6. 2024CRUXEval: A Benchmark for Code Reasoning, Understanding and Executioncited by 294
  7. 2023Magicoder: Empowering Code Generation with OSS-Instructcited by 251
  8. 2024SWE-bench Multimodal: Do AI Systems Generalize to Visual Software Domains?cited by 196
  9. 2023An Empirical Evaluation of Using Large Language Models for Automated Unit Test Generationcited by 193
  10. 2023CrossCodeEval: A Diverse and Multilingual Benchmark for Cross-File Code Completioncited by 181
  11. 2023ClassEval: A Manually-Crafted Benchmark for Evaluating LLMs on Class-level Code Generationcited by 166
  12. 2022Practical Program Repair in the Era of Large Pre-trained Language Modelscited by 133
  13. 2025ACECODER: Acing Coder RL via Automated Test-Case Synthesiscited by 89
  14. 2025SWE-Fixer: Training Open-Source LLMs for Effective and Efficient GitHub Issue Resolutioncited by 87
  15. 2023Impact of Code Language Models on Automated Program Repaircited by 85
  16. 2023No More Manual Tests? Evaluating and Improving ChatGPT for Unit Test Generationcited by 82
  17. 2022Less Training, More Repairing Please: Revisiting Automated Program Repair via Zero-shot Learningcited by 76
  18. 2024Evaluating Language Models for Efficient Code Generationcited by 69
  19. 2023Copiloting the Copilots: Fusing Large Language Models with Completion Engines for Automated Program Repaircited by 55
  20. 2024TestGenEval: A Real World Unit Test Generation and Test Completion Benchmarkcited by 49
  21. 2024Fewer Truncations Improve Language Modelingcited by 33

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 120; 2026: 126.

246 papers.

Sort:

By area:

  1. 2025Group-in-Group Policy Optimization for LLM Agent Trainingcited by 442
  2. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  3. 2025SWE-smith: Scaling Data for Software Engineering Agentscited by 255
  4. 2025TTRL: Test-Time Reinforcement Learningcited by 253
  5. 2025The Landscape of Agentic Reinforcement Learning for LLMs: A Surveycited by 195
  6. 2025A Survey of Reinforcement Learning for Large Reasoning Modelscited by 175
  7. 2025R2E-Gym: Procedural Environments and Hybrid Verifiers for Scaling Open-Weights SWE Agentscited by 140
  8. 2025OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automationcited by 129
  9. 2025A Survey of Frontiers in LLM Reasoning: Inference Scaling, Learning to Reason, and Agentic Systemscited by 108
  10. 2025Self-Evolving Curriculum for LLM Reasoningcited by 107
  11. 2025Online Difficulty Filtering for Reasoning Oriented Reinforcement Learningcited by 100
  12. 2025Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly?cited by 91
38 more
  1. 2025CWM: An Open-Weights LLM for Research on Code Generation with World Modelscited by 87
  2. 2025Dream-Coder 7B: An Open Diffusion Language Model for Codecited by 81
  3. 2026CaP-X: A Framework for Benchmarking and Improving Coding Agents for Robot Manipulationcited by 72
  4. 2025SEC-bench: Automated Benchmarking of LLM Agents on Real-World Software Security Taskscited by 69
  5. 2025SWE-Exp: Experience-Driven Software Issue Resolutioncited by 64
  6. 2026MOPD: Multi-Teacher On-Policy Distillation for Capability Integration in LLM Post-Trainingcited by 60
  7. 2025Reasoning-SQL: Reinforcement Learning with SQL Tailored Partial Rewards for Reasoning-Enhanced Text-to-SQLcited by 59
  8. 2025CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learningcited by 59
  9. 2025Stop Summation: Min-Form Credit Assignment Is All Process Reward Model Needs for Reasoningcited by 49
  10. 2025SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarioscited by 49
  11. 2026A Survey of Agentic Reasoning for Large Language Models: Towards Recursively Self-Improving and Collective Agentscited by 47
  12. 2025Agentic Reinforcement Learning with Implicit Step Rewardscited by 45
  13. 2025Nemotron-Cascade: Scaling Cascaded Reinforcement Learning for General-Purpose Reasoning Modelscited by 45
  14. 2025ContextAgent: Context-Aware Proactive LLM Agents with Open-World Sensory Perceptionscited by 44
  15. 2025Harnessing Uncertainty: Entropy-Modulated Policy Gradients for Long-Horizon LLM Agentscited by 43
  16. 2025Agent-RLVR: Training Software Engineering Agents via Guidance and Environment Rewardscited by 41
  17. 2025Agentic Software Engineering: Foundational Pillars and a Research Roadmapcited by 40
  18. 2025Large Language Models for Planning: A Comprehensive and Systematic Surveycited by 38
  19. 2025Context as a Tool: Context Management for Long-Horizon SWE-Agentscited by 38
  20. 2025LlamaRL: A Distributed Asynchronous Reinforcement Learning Framework for Efficient Large-scale LLM Trainingcited by 37
  21. 2026Hierarchy-of-Groups Policy Optimization for Long-Horizon Agentic Taskscited by 37
  22. 2025Z1: Efficient Test-time Scaling with Codecited by 35
  23. 2025Kimi-Dev: Agentless Training as Skill Prior for SWE-Agentscited by 35
  24. 2025VChain: Chain-of-Visual-Thought for Reasoning in Video Generationcited by 31
  25. 2025RAG-RL: Advancing Retrieval-Augmented Generation via RL and Curriculum Learningcited by 30
  26. 2025Beyond Verifiable Rewards: Scaling Reinforcement Learning for Language Models to Unverifiable Datacited by 29
  27. 2025100 Days After DeepSeek-R1: A Survey on Replication Studies and More Directions for Reasoning Language Modelscited by 29
  28. 2025PipelineRL: Faster On-policy Reinforcement Learning for Long Sequence Generationcited by 29
  29. 2025BugPilot: Complex Bug Generation for Efficient Learning of SWE Skillscited by 28
  30. 2025Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystemcited by 28
  31. 2025SPELL: Self-Play Reinforcement Learning for Evolving Long-Context Language Modelscited by 27
  32. 2025Thinking Longer, Not Larger: Enhancing Software Engineering Agents via Scaling Test-Time Computecited by 25
  33. 2025SWE-RM: Execution-free Feedback For Software Engineering Agentscited by 24
  34. 2025Issue Localization via LLM-Driven Iterative Code Graph Searchingcited by 22
  35. 2025SWE-Synth: Synthesizing Verifiable Bug-Fix Data to Enable Large Language Models in Resolving Real-World Bugscited by 22
  36. 2025Cyber-Zero: Training Cybersecurity Agents without Runtimecited by 22
  37. 2026Agentic Rubrics as Contextual Verifiers for SWE Agentscited by 22
  38. 2025Entropy-Gated Branching for Efficient Test-Time Reasoningcited by 21

50 most cited of 246 All 246 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain Perspective
  2. 2026SWE-rebench V2: Language-Agnostic SWE Task Collection at Scale
  3. 2025ReTool: Reinforcement Learning for Strategic Tool Use in LLMs
  4. 2025Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training
  5. 2025Part I: Tricks or Traps? A Deep Dive into RL for LLM Reasoning
  6. 2024RLSF: Fine-tuning LLMs via Symbolic Feedback
  7. 2025R-Log: Incentivizing Log Analysis Capability in LLMs via Reasoning-based Reinforcement Learning
  8. 2025R-Search: Empowering LLM Reasoning with Search via Multi-Reward Reinforcement Learning