一作ACL 2026Findings· Findings of ACL
Efficient Hallucination Detection in Automatic Code Generation
Georgii Andriushchenko · Roman Garaev · Lyudmila Rvanova · Artem Shelmanov · Vladimir V. Ivanov
摘要
Large language models (LLMs) frequently produce source code that seems correct and well-formed, yet includes hallucinated elements that cause downstream test failures. In this study, we benchmark state-of-the-art uncertainty quantification methods and existing baselines for the task of hallucinat...