@inproceedings{galimzianova-etal-2025-rag,
title = "From {RAG} to Reality: Coarse-Grained Hallucination Detection via {NLI} Fine-Tuning",
author = "Galimzianova, Daria and
Boriskin, Aleksandr and
Arshinov, Grigory",
editor = "Ghosal, Tirthankar and
Mayr, Philipp and
Singh, Amanpreet and
Naik, Aakanksha and
Rehm, Georg and
Freitag, Dayne and
Li, Dan and
Schimmler, Sonja and
De Waard, Anita",
booktitle = "Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.sdp-1.34/",
doi = "10.18653/v1/2025.sdp-1.34",
pages = "353--359",
ISBN = "979-8-89176-265-7",
abstract = "We present our submission to SciHal Subtask 1: coarse-grained hallucination detection for scientific question answering. We frame hallucination detection as an NLI-style three-way classification (entailment, contradiction, unverifiable) and show that simple fine-tuning of NLI-adapted encoder models on task data outperforms more elaborate feature-based pipelines and large language model prompting. In particular, DeBERTa-V3-large, a model pretrained on five diverse NLI corpora, achieves the highest weighted F1 on the public leaderboard. We additionally explore a pipeline combining joint claim{--}reference embeddings and NLI softmax probabilities fed into a classifier, but find its performance consistently below direct encoder fine-tuning. Our findings demonstrate that, for reference-grounded hallucination detection, targeted encoder fine-tuning remains the most accurate and efficient approach."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="galimzianova-etal-2025-rag">
<titleInfo>
<title>From RAG to Reality: Coarse-Grained Hallucination Detection via NLI Fine-Tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Daria</namePart>
<namePart type="family">Galimzianova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aleksandr</namePart>
<namePart type="family">Boriskin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Grigory</namePart>
<namePart type="family">Arshinov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tirthankar</namePart>
<namePart type="family">Ghosal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Philipp</namePart>
<namePart type="family">Mayr</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amanpreet</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aakanksha</namePart>
<namePart type="family">Naik</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Georg</namePart>
<namePart type="family">Rehm</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dayne</namePart>
<namePart type="family">Freitag</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sonja</namePart>
<namePart type="family">Schimmler</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anita</namePart>
<namePart type="family">De Waard</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-265-7</identifier>
</relatedItem>
<abstract>We present our submission to SciHal Subtask 1: coarse-grained hallucination detection for scientific question answering. We frame hallucination detection as an NLI-style three-way classification (entailment, contradiction, unverifiable) and show that simple fine-tuning of NLI-adapted encoder models on task data outperforms more elaborate feature-based pipelines and large language model prompting. In particular, DeBERTa-V3-large, a model pretrained on five diverse NLI corpora, achieves the highest weighted F1 on the public leaderboard. We additionally explore a pipeline combining joint claim–reference embeddings and NLI softmax probabilities fed into a classifier, but find its performance consistently below direct encoder fine-tuning. Our findings demonstrate that, for reference-grounded hallucination detection, targeted encoder fine-tuning remains the most accurate and efficient approach.</abstract>
<identifier type="citekey">galimzianova-etal-2025-rag</identifier>
<identifier type="doi">10.18653/v1/2025.sdp-1.34</identifier>
<location>
<url>https://aclanthology.org/2025.sdp-1.34/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>353</start>
<end>359</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T From RAG to Reality: Coarse-Grained Hallucination Detection via NLI Fine-Tuning
%A Galimzianova, Daria
%A Boriskin, Aleksandr
%A Arshinov, Grigory
%Y Ghosal, Tirthankar
%Y Mayr, Philipp
%Y Singh, Amanpreet
%Y Naik, Aakanksha
%Y Rehm, Georg
%Y Freitag, Dayne
%Y Li, Dan
%Y Schimmler, Sonja
%Y De Waard, Anita
%S Proceedings of the Fifth Workshop on Scholarly Document Processing (SDP 2025)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-265-7
%F galimzianova-etal-2025-rag
%X We present our submission to SciHal Subtask 1: coarse-grained hallucination detection for scientific question answering. We frame hallucination detection as an NLI-style three-way classification (entailment, contradiction, unverifiable) and show that simple fine-tuning of NLI-adapted encoder models on task data outperforms more elaborate feature-based pipelines and large language model prompting. In particular, DeBERTa-V3-large, a model pretrained on five diverse NLI corpora, achieves the highest weighted F1 on the public leaderboard. We additionally explore a pipeline combining joint claim–reference embeddings and NLI softmax probabilities fed into a classifier, but find its performance consistently below direct encoder fine-tuning. Our findings demonstrate that, for reference-grounded hallucination detection, targeted encoder fine-tuning remains the most accurate and efficient approach.
%R 10.18653/v1/2025.sdp-1.34
%U https://aclanthology.org/2025.sdp-1.34/
%U https://doi.org/10.18653/v1/2025.sdp-1.34
%P 353-359
Markdown (Informal)
[From RAG to Reality: Coarse-Grained Hallucination Detection via NLI Fine-Tuning](https://aclanthology.org/2025.sdp-1.34/) (Galimzianova et al., sdp 2025)
ACL