Source code for ICML 2025 paper Steer LLM Latents for Hallucination Detection by Seongheon Park, Xuefeng Du, Min-Hsuan Yeh, Haobo Wang, and Yixuan Li

conda env create -f tsv.yml

Generate responses for each question to construct an unlabeled QA dataset in the wild.

bash gen.sh

Generate BLEURT score for each QA pair

bash gt.sh

Train TSV for hallucination detection.

bash train.sh

@inproceedings{
park2025steer,
title={Steer {LLM} Latents for Hallucination Detection},
author={Seongheon Park and Xuefeng Du and Min-Hsuan Yeh and Haobo Wang and Yixuan Li},
booktitle={Forty-second International Conference on Machine Learning},
year={2025}
}

We gratefully acknowledge HaloScope, ITI, and ICV for their inspiring ideas and open-source contributions, which served as valuable foundations for this work.