Interleaving Static Analysis and LLM Prompting for Error-Specification Inference

Explore a 23-minute video presentation from the SOAP 2024 conference that introduces a novel approach combining static program analysis with Large Language Models (LLMs). Learn how researchers from the University of California at Davis interleave static analyzer calls with LLM queries to enhance program analysis, particularly for error-specification inference in C systems code. Discover how this method significantly improves recall and F1-scores while maintaining precision when compared to existing state-of-the-art techniques. Gain insights into the application of this approach on real-world C programs like MbedTLS and zlib, and understand its potential impact on program understanding and error-handling bug detection.