Get Started
Home
Topics
Search
Library
Research questionHow can we detect when LLM-decompiled code diverges from original behavior or erases disclosed vulnerabilities?Compilability and shipped input/output tests can pass even when reconstructed code behaves differently on legitimate inputs or no longer reproduces a disclosed crash. This complicates the use of decompiled code for vulnerability detection and malware analysis.
AI
Code Generation & Program Synthesis
Evaluation & Benchmarks
Technology
Latest papersRecent research connected to this question, newest first.When LLM Decompilers Recompile More and Preserve LessThe evidence concerns LLM-based decompilers producing C from compiled machine code. It covers established decompilation corpora, 300 GitHub library functions, and 287 CVE-grounded functions across eight systems and nine configurations, comparing recompilability with behavioral agreement and crash preservation on generated test inputs; it does not establish correctness for all possible inputs.research paper · Sep 4, 2026
Related questions
How can LLMs generate functionally correct code without introducing security vulnerabilities?How can black-box LLMs resist jailbreaks without weight access or retraining while preserving benign-query utility?How can LLM sandbox security remain reliable when linguistic monitoring misrepresents internal computation?How can LLM-based automated program repair detect patches and artifacts unsupported by repair evidence?