Detecting and Correcting Hallucinations in LLM-Generated Code via Deterministic AST Analysis

Dipin Khati, Daniel Rodriguez-Cardenas, Paul Pantzer +1

Large Language Models (LLMs) for code generation boost productivity but frequently introduce Knowledge Conflicting Hallucinations (KCHs), subtle, semantic errors, such as non-exist…