Hallucination-Resilient LLM-Driven Sound and Tunable Static Analysis: A Case of Higher-Order Control-Flow Analysis
Year: 2025
Type: conference-paper
Abstract: We argue that soundness remains essential for LLM-driven static analysis and discuss hallucination-resilient approaches in combining LLMs with static analysis that ensure soundness while improving precision. We propose to use LLMs as a way of meta-analysis and investigate this approach in higher-order control-flow analysis, building on the abstracting abstract machine framework and delegating abstract address allocation to an LLM. Our analyzer LLMAAM maintains soundness regardless of LLM behavior, while adaptively tuning analysis precision. We report promising preliminary results and outline broader opportunities for sound LLM-driven analysis.
Authors: Guannan Wei, Zhuo Zhang, Caterina Urban
Institutions: Tufts University, Institut national de recherche en sciences et technologies du numérique, Université Paris Sciences et Lettres, École Normale Supérieure - PSL, Shanghai Jiao Tong University, Columbia University, Département d'Informatique, ANTIQUE: Analyse Statique par Interprétation Abstraite
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Funder: Agence Nationale de la Recherche
OpenAlex ID: W4414827940