@inproceedings{chen-etal-2026-generation,
title = "Generation-Score Consistency: Pushing {LLM}s to Generate Semantically Valid Reasoning Scores for Misinformation Detection",
author = "Chen, Yidu and
Ma, Bo and
Yang, Yating and
Zhou, Xi and
Wang, Zhen and
Ahmat, Ahtamjan and
Abdureyim, Dilxat and
Anwar, Azmat",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.973/",
pages = "22435--22455",
ISBN = "979-8-89176-459-0",
abstract = "Large language models (LLMs) possess extensive world knowledge and have demonstrated strong potential for misinformation detection. A growing line of recent work prompts LLMs to produce structured reasoning before issuing a veracity judgment. However, we contend that such structured outputs do not spontaneously support the veracity decision. The emitted scores can satisfy the output schema while contributing little to veracity discrimination. Since annotating full reasoning paths is prohibitively costly for veracity tasks, we propose a Generation-Score Consistency framework that supervises through structured dimension scores and steers the LLM toward functional reasoning. Specifically, we train a compact score-to-label verifier on a small set of expert-annotated reasoning data, freeze it, and during supervised fine-tuning use a trainable score head to align the score-token hidden states of the LLM with the human-calibrated mapping. Introducing structured reasoning on its own costs accuracy relative to label-only supervision, while the consistency term recovers that cost and keeps the emitted scores coupled to the veracity decision. Controlled experiments on multimodal misinformation benchmarks show that the consistency term strengthens both veracity prediction and the discriminability of the emitted scores. Code and data are available at https://github.com/yun-nlp/Reasoning-Scores-for-Misinformation-Detection-EMNLP2026."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-generation">
<titleInfo>
<title>Generation-Score Consistency: Pushing LLMs to Generate Semantically Valid Reasoning Scores for Misinformation Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yidu</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bo</namePart>
<namePart type="family">Ma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yating</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xi</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhen</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ahtamjan</namePart>
<namePart type="family">Ahmat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dilxat</namePart>
<namePart type="family">Abdureyim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Azmat</namePart>
<namePart type="family">Anwar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Large language models (LLMs) possess extensive world knowledge and have demonstrated strong potential for misinformation detection. A growing line of recent work prompts LLMs to produce structured reasoning before issuing a veracity judgment. However, we contend that such structured outputs do not spontaneously support the veracity decision. The emitted scores can satisfy the output schema while contributing little to veracity discrimination. Since annotating full reasoning paths is prohibitively costly for veracity tasks, we propose a Generation-Score Consistency framework that supervises through structured dimension scores and steers the LLM toward functional reasoning. Specifically, we train a compact score-to-label verifier on a small set of expert-annotated reasoning data, freeze it, and during supervised fine-tuning use a trainable score head to align the score-token hidden states of the LLM with the human-calibrated mapping. Introducing structured reasoning on its own costs accuracy relative to label-only supervision, while the consistency term recovers that cost and keeps the emitted scores coupled to the veracity decision. Controlled experiments on multimodal misinformation benchmarks show that the consistency term strengthens both veracity prediction and the discriminability of the emitted scores. Code and data are available at https://github.com/yun-nlp/Reasoning-Scores-for-Misinformation-Detection-EMNLP2026.</abstract>
<identifier type="citekey">chen-etal-2026-generation</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.973/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>22435</start>
<end>22455</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Generation-Score Consistency: Pushing LLMs to Generate Semantically Valid Reasoning Scores for Misinformation Detection
%A Chen, Yidu
%A Ma, Bo
%A Yang, Yating
%A Zhou, Xi
%A Wang, Zhen
%A Ahmat, Ahtamjan
%A Abdureyim, Dilxat
%A Anwar, Azmat
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F chen-etal-2026-generation
%X Large language models (LLMs) possess extensive world knowledge and have demonstrated strong potential for misinformation detection. A growing line of recent work prompts LLMs to produce structured reasoning before issuing a veracity judgment. However, we contend that such structured outputs do not spontaneously support the veracity decision. The emitted scores can satisfy the output schema while contributing little to veracity discrimination. Since annotating full reasoning paths is prohibitively costly for veracity tasks, we propose a Generation-Score Consistency framework that supervises through structured dimension scores and steers the LLM toward functional reasoning. Specifically, we train a compact score-to-label verifier on a small set of expert-annotated reasoning data, freeze it, and during supervised fine-tuning use a trainable score head to align the score-token hidden states of the LLM with the human-calibrated mapping. Introducing structured reasoning on its own costs accuracy relative to label-only supervision, while the consistency term recovers that cost and keeps the emitted scores coupled to the veracity decision. Controlled experiments on multimodal misinformation benchmarks show that the consistency term strengthens both veracity prediction and the discriminability of the emitted scores. Code and data are available at https://github.com/yun-nlp/Reasoning-Scores-for-Misinformation-Detection-EMNLP2026.
%U https://aclanthology.org/2026.emnlp-main.973/
%P 22435-22455
Markdown (Informal)
[Generation-Score Consistency: Pushing LLMs to Generate Semantically Valid Reasoning Scores for Misinformation Detection](https://aclanthology.org/2026.emnlp-main.973/) (Chen et al., EMNLP 2026)
ACL
- Yidu Chen, Bo Ma, Yating Yang, Xi Zhou, Zhen Wang, Ahtamjan Ahmat, Dilxat Abdureyim, and Azmat Anwar. 2026. Generation-Score Consistency: Pushing LLMs to Generate Semantically Valid Reasoning Scores for Misinformation Detection. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 22435–22455, Budapest, Hungary. Association for Computational Linguistics.