@inproceedings{huo-etal-2024-mmneuron,
title = "{MMN}euron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model",
author = "Huo, Jiahao and
Yan, Yibo and
Hu, Boren and
Yue, Yutao and
Hu, Xuming",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.emnlp-main.387/",
doi = "10.18653/v1/2024.emnlp-main.387",
pages = "6801--6816",
abstract = "Projecting visual features into word embedding space has become a significant fusion strategy adopted by Multimodal Large Language Models (MLLMs). However, its internal mechanisms have yet to be explored. Inspired by multilingual research, we identify domain-specific neurons in multimodal large language models. Specifically, we investigate the distribution of domain-specific neurons and the mechanism of how MLLMs process features from diverse domains. Furthermore, we propose a three-stage framework for language model modules in MLLMs when handling projected image features, and verify this hypothesis using logit lens. Extensive experiments indicate that while current MLLMs exhibit Visual Question Answering (VQA) capability, they may not fully utilize domain-specific information. Manipulating domain-specific neurons properly will result in a 10{\%} change of accuracy at most, shedding light on the development of cross-domain, all-encompassing MLLMs in the future. The source code is available at https://anonymous.4open.science/r/MMNeuron."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huo-etal-2024-mmneuron">
<titleInfo>
<title>MMNeuron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jiahao</namePart>
<namePart type="family">Huo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yibo</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Boren</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yutao</namePart>
<namePart type="family">Yue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuming</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yaser</namePart>
<namePart type="family">Al-Onaizan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Nung</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Projecting visual features into word embedding space has become a significant fusion strategy adopted by Multimodal Large Language Models (MLLMs). However, its internal mechanisms have yet to be explored. Inspired by multilingual research, we identify domain-specific neurons in multimodal large language models. Specifically, we investigate the distribution of domain-specific neurons and the mechanism of how MLLMs process features from diverse domains. Furthermore, we propose a three-stage framework for language model modules in MLLMs when handling projected image features, and verify this hypothesis using logit lens. Extensive experiments indicate that while current MLLMs exhibit Visual Question Answering (VQA) capability, they may not fully utilize domain-specific information. Manipulating domain-specific neurons properly will result in a 10% change of accuracy at most, shedding light on the development of cross-domain, all-encompassing MLLMs in the future. The source code is available at https://anonymous.4open.science/r/MMNeuron.</abstract>
<identifier type="citekey">huo-etal-2024-mmneuron</identifier>
<identifier type="doi">10.18653/v1/2024.emnlp-main.387</identifier>
<location>
<url>https://aclanthology.org/2024.emnlp-main.387/</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>6801</start>
<end>6816</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MMNeuron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model
%A Huo, Jiahao
%A Yan, Yibo
%A Hu, Boren
%A Yue, Yutao
%A Hu, Xuming
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F huo-etal-2024-mmneuron
%X Projecting visual features into word embedding space has become a significant fusion strategy adopted by Multimodal Large Language Models (MLLMs). However, its internal mechanisms have yet to be explored. Inspired by multilingual research, we identify domain-specific neurons in multimodal large language models. Specifically, we investigate the distribution of domain-specific neurons and the mechanism of how MLLMs process features from diverse domains. Furthermore, we propose a three-stage framework for language model modules in MLLMs when handling projected image features, and verify this hypothesis using logit lens. Extensive experiments indicate that while current MLLMs exhibit Visual Question Answering (VQA) capability, they may not fully utilize domain-specific information. Manipulating domain-specific neurons properly will result in a 10% change of accuracy at most, shedding light on the development of cross-domain, all-encompassing MLLMs in the future. The source code is available at https://anonymous.4open.science/r/MMNeuron.
%R 10.18653/v1/2024.emnlp-main.387
%U https://aclanthology.org/2024.emnlp-main.387/
%U https://doi.org/10.18653/v1/2024.emnlp-main.387
%P 6801-6816
Markdown (Informal)
[MMNeuron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model](https://aclanthology.org/2024.emnlp-main.387/) (Huo et al., EMNLP 2024)
ACL