Artifact - A Fine-Grained Program Analysis of LLM-Generated Code via Line-Level Uncertainty Signals

This artifact contains datasets, code, and evaluation scripts for line-level hallucination detection in LLM-generated code. It includes annotated benchmarks, implementations of multiple uncertainty quantification methods, and tools to reproduce the experiments and analyses in our paper “A Fine-Grained Program Analysis of LLM-Generated Code via Line-Level Uncertainty Signals.”