mlsanityData to 14 Oct 2026

Noisy Test-Time Reinforcement Learning for Code LLMs

arXiv 2609.32172 ↗First posted 26 Sep 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) have demonstrated remarkable performance across various code-related tasks. However, unlike carefully curated datasets that are typically high-quality and error-free, real-world user instructions are often vague and error-prone, posing significant challenges to the robustness of code LLMs. Furthermore, robustness-oriented fine-tuning relies on paired clean-noisy samples, which are costly to curate and require sophisticated noisy simulation techniques. To address these challenges, we propose the Noisy Test-time Reinforcement Learning framework (NTRL-Code), which enables robust self-evolution of code LLMs using only unlabeled noisy data during the testing stage. Specifically, NTRL-Code uses conservative self-denoising to obtain a cleaner semantic anchor for target estimation, and employs an abstract-syntax-tree (AST)-based structural aggregation mechanism to estimate a proxy target from multiple candidate programs. The policy is then optimized on the original noisy prompts with a hybrid reward that combines format validity, code similarity, and anti-repetition signals. Extensive experiments on three benchmarks, each incorporating character-level, word-level, and paragraph-level perturbations, demonstrate that NTRL-Code yields robust and consistent improvements, stabilizing the predictions of various base models. Our code is available at https://github.com/Xikai97/NTRL-Code.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 24 of its 42 references that are in the map. First posted on arXiv 26 Sep 2026.

Builds on

24 of 42 references.

References · 24

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  3. 2021Program Synthesis with Large Language Modelscited by 3,637
  4. 2024Qwen2.5-Coder Technical Reportcited by 1,203
  5. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,181
  6. 2024BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructionscited by 546
  7. 2023PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Promptscited by 226
  8. 2025Right Question is Already Half the Answer: Fully Unsupervised LLM Reasoning Incentivizationcited by 128
  9. 2024Improving the Robustness of Large Language Models via Consistency Alignmentcited by 36
  10. 2016A note on the triangle inequality for the Jaccard distancecited by 32
  11. 2024Robustness, Security, Privacy, Explainability, Efficiency, and Usability of Large Language Models for Codecited by 28
  12. 2024HumanEvalComm: Benchmarking the Communication Competence of Code Generation for LLMs and LLM Agentcited by 24
12 more
  1. 2023Certified Robustness for Large Language Models with Self-Denoisingcited by 23
  2. 2025Dynamic Benchmarking of Reasoning Capabilities in Code Large Language Models Under Data Contaminationcited by 23
  3. 2024NLPerturbator: Studying the Robustness of Code LLMs to Natural Language Variationscited by 20
  4. 2025Adversarial Training of Reward Modelscited by 18
  5. 2024Towards Understanding the Robustness of LLM-based Evaluations under Perturbationscited by 10
  6. 2024On the Limitations of Embedding Based Methods for Measuring Functional Correctness for Code Generationcited by 6
  7. 2025SAGE: A Realistic Benchmark for Semantic Understandingcited by 6
  8. 2025RobuNFR: Evaluating the Robustness of Large Language Models on Non-Functional Requirements Aware Code Generationcited by 5
  9. 2025CREME: Robustness Enhancement of Code LLMs via Layer-Aware Model Editingcited by 5
  10. 2025Code Roulette: How Prompt Variability Affects LLM Code Generationcited by 4
  11. 2024On the Adversarial Robustness of Instruction-Tuned Large Language Models for Codecited by 3
  12. 2025Adversarial Attack Classification and Robustness Testing for Large Language Models for Codecited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025TTRL: Test-Time Reinforcement Learning
  2. 2026CodeScaler: Scaling Code LLM Training and Test-Time Inference via Reward Models
  3. 2024RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning
  4. 2026Two-Stage Reinforcement Learning for Sound and Adversarial Test Generation in Code LLMs
  5. 2024Reinforcement Learning from Automatic Feedback for High-Quality Unit Test Generation
  6. 2026Adaptive Decoding via Test-Time Policy Learning for Self-Improving Generation
  7. 2025Rethinking Verification for LLM Code Generation: From Generation to Testing
  8. 2026Scaling with Confidence: Calibrating Confidence of LLMs for Adaptive Test Time Scaling