Summary: This repository curates research papers on using Large Language Models (LLMs) for software vulnerability detection. It targets researchers and practitioners in cybersecurity and software engineering, providing a comprehensive overview of state-of-the-art techniques and evaluation frameworks in this rapidly advancing field. The primary benefit is a centralized resource for understanding LLM applications in code security.

How It Works: The collection highlights diverse approaches to applying LLMs in vulnerability detection. Papers explore context-aware methods using Code Property Graphs, multi-modal contrastive learning, and reinforcement learning for vulnerability reasoning. Many works focus on benchmarking LLM performance against real-world code and evaluating their reasoning capabilities. Methodologies range from prompt engineering and mixture-of-experts tuning to integrating LLMs with static analysis, data-flow analysis, and program synthesis for detection and repair. The core advantage lies in harnessing LLMs' code comprehension and pattern recognition for enhanced security flaw identification.

Quick Start & Requirements: This repository is a curated list of research papers, not an executable project. Installation instructions or specific requirements are not applicable.

Highlighted Details:

• Extensive Benchmarking: A significant portion focuses on evaluating LLM capabilities for vulnerability detection across various contexts (e.g., SecVulEval, CVE-Bench, LLM4Vuln).
• Diverse Methodologies: Covers fine-tuning, prompt engineering, multi-modal learning, graph-based approaches, and LLM-integrated static analysis.
• Practical Application Focus: Research spans smart contract security, C/C++ vulnerability detection, and IDE integration.
• Emerging Trends: Majority of papers from 2024/2025, indicating a highly active area with emphasis on LLM limitations and robustness.

Maintenance & Community: The README indicates an "Automated daily capture and update of Arxiv papers" workflow, suggesting automated maintenance for the paper list. No direct community interaction channels are mentioned.

Licensing & Compatibility: No licensing information is provided within the README content. Compatibility for commercial use or closed-source linking cannot be determined.

Limitations & Caveats: This repository is a bibliographical resource, not an executable tool. Provided "links" are often placeholders and may not lead to accessible papers or code. The rapid pace of LLM development means the state-of-the-art shifts quickly; this list is a snapshot. Some cited works may be pre-prints or under review.