An Empirical Analysis of Static Analysis Methods for Detection and Mitigation of Code Library Hallucinations
arXiv:2604.07755v2 Announce Type: replace
Abstract: Despite extensive research, Large Language Models continue to hallucinate when generating code, particularly when using libraries. On NL-to-code benchmarks that require library use, we find that LLMs…