LLM-Enhanced Loop Invariant Synthesis for Complex Programs

Created using ChatSlide

This research explores leveraging large language models (LLMs) for generating programme invariants, specifically focusing on loop invariants critical for programme verification. Traditional methods face significant challenges, especially in complex programmes. This study addresses these limitations by introducing a novel LLM-based approach. The methodology includes extracting and analysing abstract syntax trees (AST) and control flow graphs (CFG), applying a branch-based decomposition...