@inproceedings{d-mishra-2026-grain,
title = "{GRAIN}: Gradient-Based Interaction Graph of Neurons for Explainable Misinformation Detection",
author = "D, Rathod Darshan and
Mishra, Rahul",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.183/",
pages = "3655--3670",
ISBN = "979-8-89176-444-6",
abstract = "Prior work on misinformation detection often applies large language models (LLMs) in a black-box manner, relying on knowledge implicitly encoded in model parameters or on pre-hoc and post-hoc reasoning techniques. However, the opaque nature of LLMs limits access to their internal decision-making processes, and existing approaches largely overlook the rich signals embedded within their hidden layers, instead focusing primarily on surface-level generative capabilities. To address these limitations, we propose GRAIN, an explainable misinformation detection framework that explicitly models neuron{--}neuron interaction graphs using dominancy scores. Experimental results across multiple benchmark datasets show that GRAIN consistently outperforms state-of-the-art baselines while providing interpretable insights into the neuronal patterns and cues that distinguish misinformation from factual content. Furthermore, we introduce NNI-IG, a GRAIN-guided variant of Integrated Gradients that leverages critical neurons to produce sharper and more focused token-level attributions, further enhancing model interpretability."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="d-mishra-2026-grain">
<titleInfo>
<title>GRAIN: Gradient-Based Interaction Graph of Neurons for Explainable Misinformation Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rathod</namePart>
<namePart type="given">Darshan</namePart>
<namePart type="family">D</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rahul</namePart>
<namePart type="family">Mishra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Prior work on misinformation detection often applies large language models (LLMs) in a black-box manner, relying on knowledge implicitly encoded in model parameters or on pre-hoc and post-hoc reasoning techniques. However, the opaque nature of LLMs limits access to their internal decision-making processes, and existing approaches largely overlook the rich signals embedded within their hidden layers, instead focusing primarily on surface-level generative capabilities. To address these limitations, we propose GRAIN, an explainable misinformation detection framework that explicitly models neuron–neuron interaction graphs using dominancy scores. Experimental results across multiple benchmark datasets show that GRAIN consistently outperforms state-of-the-art baselines while providing interpretable insights into the neuronal patterns and cues that distinguish misinformation from factual content. Furthermore, we introduce NNI-IG, a GRAIN-guided variant of Integrated Gradients that leverages critical neurons to produce sharper and more focused token-level attributions, further enhancing model interpretability.</abstract>
<identifier type="citekey">d-mishra-2026-grain</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.183/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>3655</start>
<end>3670</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T GRAIN: Gradient-Based Interaction Graph of Neurons for Explainable Misinformation Detection
%A D, Rathod Darshan
%A Mishra, Rahul
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F d-mishra-2026-grain
%X Prior work on misinformation detection often applies large language models (LLMs) in a black-box manner, relying on knowledge implicitly encoded in model parameters or on pre-hoc and post-hoc reasoning techniques. However, the opaque nature of LLMs limits access to their internal decision-making processes, and existing approaches largely overlook the rich signals embedded within their hidden layers, instead focusing primarily on surface-level generative capabilities. To address these limitations, we propose GRAIN, an explainable misinformation detection framework that explicitly models neuron–neuron interaction graphs using dominancy scores. Experimental results across multiple benchmark datasets show that GRAIN consistently outperforms state-of-the-art baselines while providing interpretable insights into the neuronal patterns and cues that distinguish misinformation from factual content. Furthermore, we introduce NNI-IG, a GRAIN-guided variant of Integrated Gradients that leverages critical neurons to produce sharper and more focused token-level attributions, further enhancing model interpretability.
%U https://aclanthology.org/2026.findings-aacl.183/
%P 3655-3670
Markdown (Informal)
[GRAIN: Gradient-Based Interaction Graph of Neurons for Explainable Misinformation Detection](https://aclanthology.org/2026.findings-aacl.183/) (D & Mishra, Findings 2026)
ACL