The field of automated program repair and code analysis is rapidly evolving, with a focus on improving the efficiency and effectiveness of large language models (LLMs) in identifying and fixing bugs. Recent developments have explored the use of internal reflection and external feedback mechanisms to refine patch generation, as well as the application of LLMs to firmware security analysis and RTL repair. Notable advancements include the integration of code summarization as a diagnostic tool for program repair and the development of novel frameworks for reliable RTL repair and path-aware CLI fuzzing. Noteworthy papers include:

• TokenRepair, which enhances APR by integrating internal reflection for localizing faulty tokens and external feedback for quality-aware patch refinement, achieving state-of-the-art repair performance.
• FIRMHIVE, which enables LLMs to act as autonomous firmware security analysts, performing deeper and broader cross-file exploration and identifying more vulnerabilities than existing tools.
• R3A, which proposes a reliable RTL repair framework with multi-agent fault localization and stochastic tree-of-thoughts patch generation, fixing 90.6% of bugs in the RTL-repair dataset.
• PILOT, which designs a novel path-guided, iterative LLM-orchestrated testing framework for CLI fuzzing, achieving higher coverage and discovering 51 zero-day vulnerabilities.