mlsanityData to 7 Oct 2026

If It's Not Buggy, Don't Fix It: On the Dynamics of Iterative Bug-fixing with LLMs

arXiv 2609.10123 ↗First posted 9 Sep 2026cs.SE cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) have become ubiquitous in software development, with LLM-based automated program repair tools increasingly used during code review. In this report, we explore the iterative blind use of LLMs as bug-fixers. Across multiple models and repair environments, we find that LLMs consistently claim to detect bugs in entirely bug-free programs while the rate of repair of buggy programs is less than that of the damage to correct programs. We also explore the long-term dynamics of this iterative process, and find that this frequently reaches a pseudo-bug-fixing cycle where the same changes are added and removed again ad infinitum. Lastly, via mechanistic probing, we unveil the existence of a steering vector which controls the editing propensity, suggesting that LLMs have an internal representation of ``buggy code", and that this representation is what is falsely activated to induce pseudo-bug fixing. These results provide insight towards the dynamics of fully autonomous bug-fixing systems, as well as stopping conditions under ambiguous goals.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 34 of its 40 references that are in the map. First posted on arXiv 9 Sep 2026.

Builds on

34 of 40 references.

References · 34

  1. 2024Qwen2.5 Technical Reportcited by 4,713
  2. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  3. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  4. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  5. 2024SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineeringcited by 1,628
  6. 2023The Linear Representation Hypothesis and the Geometry of Large Language Modelscited by 679
  7. 2023The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetscited by 673
  8. 2025LLMs Get Lost In Multi-Turn Conversationcited by 400
  9. 2025Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approachcited by 372
  10. 2025SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolutioncited by 246
  11. 2024RepairAgent: An Autonomous, LLM-Based Agent for Program Repaircited by 225
  12. 2025AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoderscited by 199
22 more
  1. 2023Is Self-Repair a Silver Bullet for Code Generation?cited by 191
  2. 2022Do Users Write More Insecure Code with AI Assistants?cited by 186
  3. 2025Understanding Reasoning in Thinking Language Models via Steering Vectorscited by 69
  4. 2024MarsCode Agent: AI-native Automated Bug Fixingcited by 68
  5. 2025CodeContests+: High-Quality Test Case Generation for Competitive Programmingcited by 39
  6. 2026A Mechanistic Analysis of Looped Reasoning Language Modelscited by 27
  7. 2026SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Taskscited by 24
  8. 2026Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformerscited by 21
  9. 2024When LLMs Play the Telephone Game: Cultural Attractors as Conceptual Tools to Evaluate LLMs in Multi-turn Settingscited by 17
  10. 2025Unveiling Attractor Cycles in Large Language Models: A Dynamical Systems View of Successive Paraphrasingcited by 14
  11. 2025LLM as a Broken Telephone: Iterative Generation Distorts Informationcited by 12
  12. 2026LLMs Corrupt Your Documents When You Delegatecited by 9
  13. 2026Attractor States Emerge in Multi-Turn LLM Conversationscited by 6
  14. 2025ConsistencyChecker: Tree-based Evaluation of LLM Generalization Capabilitiescited by 5
  15. 2026Mutation Without Variation: Convergence Dynamics in LLM-Driven Program Evolutioncited by 4
  16. 2026The Value Axis: Language Models Encode Whether They're on the Right Trackcited by 4
  17. 2026Sparse Reward Subsystem in Large Language Modelscited by 3
  18. 2026From Restructuring to Stabilization: A Large-Scale Experiment on Iterative Code Readability Refactoring with Large Language Modelscited by 3
  19. 2026Markovian Generation Chains in Large Language Modelscited by 3
  20. 2026Do Language Models Converge to Themselves? Recursive Self-Refinement as Textual Relaxationcited by 3
  21. 2026DebugHarness: Emulating Human Dynamic Debugging for Autonomous Program Repaircited by 1
  22. 2026Looping Is Not Reliability: State-Bound Evidence and Typed Revision Contracts for Agentic Code Repaircited by 1

Built on by

Not cited yet.