Interpreting and Steering for Safe and Correct Code Generation

• URL: http://arxiv.org/abs/2608.30025v1
• Date: Sun, 30 Aug 2026 20:30:14 GMT
• Title: Interpreting and Steering for Safe and Correct Code Generation
• Abstract summary: We study how code safety-vs.vulnerability is represented or driven by components in large language models (LLMs)<n>We introduce CodeSec-Pairs, a dataset of 9,342 Python safe-and-vulnerable contrastive code pairs.<n>We propose DuoSteer, a double-steering approach that simultaneously applies safety and code-correctness steering to attention heads.
• Score: 11.245452623181398
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models (LLMs) frequently generate source code containing vulnerabilities, yet little work studies the internal mechanisms that distinguish safe from vulnerable generation in them. In this work, we systematically perform a mechanistic interpretation of LLMs, aiming at both understanding how code safety-vs-vulnerability is represented or driven by components in an LM and turning the insights into actionable steering strategies to encourage safer code generation. To this end, we introduce CodeSec-Pairs, a dataset of 9,342 Python safe-and-vulnerable contrastive code pairs, sampled from Llama-3.1-8B-Instruct. Utilizing the dataset, we explore approaches to localize layers and attention heads that relate to code safety, and further experiment with different steering strategies for inference-time vulnerability reduction. In particular, we propose DuoSteer, a double-steering approach that simultaneously applies safety and code-correctness steering to attention heads. In experiments over five vulnerability types, DuoSteer leads to an average of -26.9% vulnerability rate reduction and +7.5% functional correctness improvement, which outperforms not only other steering variants but also prompting and supervised fine-tuning baselines. The advantage also replicates on Qwen-2.5-Coder-7B-Instruct with another 2,500 contrastive pairs sampled from that model.

Related papers

• SecureForge: Finding and Preventing Vulnerabilities in LLM-Generated Code via Prompt Optimization [61.91729298584227]
  We introduce SecureForge, an automated pipeline that audits security risks of frontier models and produces auditing-informed secure system prompts.<n>SecureForge first identifies benign prompts that produce statically detectable vulnerabilities, and then amplifies them into a large synthetic prompt corpus of diverse scenarios.<n>On frontier models, SecureForge yields a statistically significant improvement in both unit test success and output security, with output vulnerabilities reduced by up to 48%.
  arXiv  Detail & Related papers  (2026-05-08T18:40:47Z)
• Structured Safety Auditing for Balancing Code Correctness and Content Safety in LLM-Generated Code [6.664505682266256]
  Large language models (LLMs) for code generation are typically evaluated on functional correctness alone.<n>We propose NLSafety-Utility Duality Score (SUDS), a metric that unifies code utility, safety adherence, and warning awareness into a single score.
  arXiv  Detail & Related papers  (2026-04-13T21:52:57Z)
• Secure Code Generation via Online Reinforcement Learning with Vulnerability Reward Model [60.60587869092729]
  Large language models (LLMs) are increasingly used in software development, yet their tendency to generate insecure code remains a major barrier to real-world deployment.<n>We propose SecCoderX, an online reinforcement learning framework for functionality-preserving secure code generation.
  arXiv  Detail & Related papers  (2026-02-07T07:42:07Z)
• ProGuard: Towards Proactive Multimodal Safeguard [48.89789547707647]
  ProGuard is a vision-language proactive guard that identifies and describes out-of-distribution (OOD) safety risks.<n>We first construct a modality-balanced dataset of 87K samples, each annotated with both binary safety labels and risk categories.<n>We then train our vision-language base model purely through reinforcement learning to achieve efficient and concise reasoning.
  arXiv  Detail & Related papers  (2025-12-29T16:13:23Z)
• DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models [50.21378052667732]
  We conduct an in-depth analysis of dLLM vulnerabilities to jailbreak attacks across two distinct dimensions: intra-step and inter-step dynamics.<n>We propose DiffuGuard, a training-free defense framework that addresses vulnerabilities through a dual-stage approach.
  arXiv  Detail & Related papers  (2025-09-29T05:17:10Z)
• Ensembling Large Language Models for Code Vulnerability Detection: An Empirical Evaluation [69.8237598448941]
  This study investigates the potential of ensemble learning to enhance the performance of Large Language Models (LLMs) in source code vulnerability detection.<n>We propose Dynamic Gated Stacking (DGS), a Stacking variant tailored for vulnerability detection.
  arXiv  Detail & Related papers  (2025-09-16T03:48:22Z)
• A Mixture of Linear Corrections Generates Secure Code [20.94236753015922]
  Large language models (LLMs) have become proficient at sophisticated code-generation tasks, yet remain ineffective at reliably detecting or avoiding code vulnerabilities.<n>We find that current LLMs encode precise internal representations that distinguish vulnerable from secure code.<n>We develop an inference-time steering technique that subtly modulates the model's token-generation probabilities through a mixture of corrections.
  arXiv  Detail & Related papers  (2025-07-13T06:27:33Z)
• HexaCoder: Secure Code Generation via Oracle-Guided Synthetic Training Data [60.75578581719921]
  Large language models (LLMs) have shown great potential for automatic code generation. Recent studies highlight that many LLM-generated code contains serious security vulnerabilities. We introduce HexaCoder, a novel approach to enhance the ability of LLMs to generate secure codes.
  arXiv  Detail & Related papers  (2024-09-10T12:01:43Z)
• Is Your AI-Generated Code Really Safe? Evaluating Large Language Models on Secure Code Generation with CodeSecEval [20.959848710829878]
  Large language models (LLMs) have brought significant advancements to code generation and code repair. However, their training using unsanitized data from open-source repositories, like GitHub, raises the risk of inadvertently propagating security vulnerabilities. We aim to present a comprehensive study aimed at precisely evaluating and enhancing the security aspects of code LLMs.
  arXiv  Detail & Related papers  (2024-07-02T16:13:21Z)

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.