Representation-based Broad Hallucination Detectors Fail to Generalize Out of Distribution

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dubanowska, Zuzanna, Żelaszczyk, Maciej, Brzozowski, Michał, Mandica, Paolo, Karpowicz, Michał
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908555886460928
author Dubanowska, Zuzanna
Żelaszczyk, Maciej
Brzozowski, Michał
Mandica, Paolo
Karpowicz, Michał
author_facet Dubanowska, Zuzanna
Żelaszczyk, Maciej
Brzozowski, Michał
Mandica, Paolo
Karpowicz, Michał
contents We critically assess the efficacy of the current SOTA in hallucination detection and find that its performance on the RAGTruth dataset is largely driven by a spurious correlation with data. Controlling for this effect, state-of-the-art performs no better than supervised linear probes, while requiring extensive hyperparameter tuning across datasets. Out-of-distribution generalization is currently out of reach, with all of the analyzed methods performing close to random. We propose a set of guidelines for hallucination detection and its evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2509_19372
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Representation-based Broad Hallucination Detectors Fail to Generalize Out of Distribution
Dubanowska, Zuzanna
Żelaszczyk, Maciej
Brzozowski, Michał
Mandica, Paolo
Karpowicz, Michał
Machine Learning
Artificial Intelligence
We critically assess the efficacy of the current SOTA in hallucination detection and find that its performance on the RAGTruth dataset is largely driven by a spurious correlation with data. Controlling for this effect, state-of-the-art performs no better than supervised linear probes, while requiring extensive hyperparameter tuning across datasets. Out-of-distribution generalization is currently out of reach, with all of the analyzed methods performing close to random. We propose a set of guidelines for hallucination detection and its evaluation.
title Representation-based Broad Hallucination Detectors Fail to Generalize Out of Distribution
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2509.19372