Saved in:
Bibliographic Details
Main Authors: Snel, Jakob, Oh, Seong Joon
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.20836
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915534519402496
author Snel, Jakob
Oh, Seong Joon
author_facet Snel, Jakob
Oh, Seong Joon
contents Large Language Models (LLMs) hallucinate, and detecting these cases is key to ensuring trust. While many approaches address hallucination detection at the response or span level, recent work explores token-level detection, enabling more fine-grained intervention. However, the distribution of hallucination signal across sequences of hallucinated tokens remains unexplored. We leverage token-level annotations from the RAGTruth corpus and find that the first hallucinated token is far more detectable than later ones. This structural property holds across models, suggesting that first hallucination tokens play a key role in token-level hallucination detection. Our code is available at https://github.com/jakobsnl/RAGTruth_Xtended.
format Preprint
id arxiv_https___arxiv_org_abs_2507_20836
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle First Hallucination Tokens Are Different from Conditional Ones
Snel, Jakob
Oh, Seong Joon
Machine Learning
Artificial Intelligence
Large Language Models (LLMs) hallucinate, and detecting these cases is key to ensuring trust. While many approaches address hallucination detection at the response or span level, recent work explores token-level detection, enabling more fine-grained intervention. However, the distribution of hallucination signal across sequences of hallucinated tokens remains unexplored. We leverage token-level annotations from the RAGTruth corpus and find that the first hallucinated token is far more detectable than later ones. This structural property holds across models, suggesting that first hallucination tokens play a key role in token-level hallucination detection. Our code is available at https://github.com/jakobsnl/RAGTruth_Xtended.
title First Hallucination Tokens Are Different from Conditional Ones
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2507.20836