@inproceedings{zhang-etal-2025-direct,
title = "Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward",
author = "Zhang, Ruohong and
Gui, Liangke and
Sun, Zhiqing and
Feng, Yihao and
Xu, Keyang and
Zhang, Yuanhan and
Fu, Di and
Li, Chunyuan and
Hauptmann, Alexander G and
Bisk, Yonatan and
Yang, Yiming",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-long.30/",
doi = "10.18653/v1/2025.naacl-long.30",
pages = "694--717",
ISBN = "979-8-89176-189-6",
abstract = "Preference modeling techniques, such as direct preference optimization (DPO), has shown effective in enhancing the generalization abilities of large language model (LLM). However, in tasks involving video instruction-following, providing informative feedback, especially for open-ended conversations, remains a significant challenge. While previous studies have explored using large multimodal models (LMMs) as reward models for guiding preference modeling, their ability to accurately assess the quality of generated responses and their alignment with video content has not been conclusively demonstrated. This paper introduces a novel framework that utilizes detailed video captions as a proxy of video content, enabling language models to incorporate this information as supporting evidence for scoring video Question Answering (QA) predictions. Our approach demonstrates robust alignment with OpenAI GPT-4V model{'}s reward mechanism, which directly takes video frames as input. Furthermore, we show that applying our reward mechanism to DPO algorithm significantly improves model performance on open-ended video QA tasks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2025-direct">
<titleInfo>
<title>Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ruohong</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liangke</namePart>
<namePart type="family">Gui</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiqing</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yihao</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Keyang</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuanhan</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Di</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chunyuan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alexander</namePart>
<namePart type="given">G</namePart>
<namePart type="family">Hauptmann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yonatan</namePart>
<namePart type="family">Bisk</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiming</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-189-6</identifier>
</relatedItem>
<abstract>Preference modeling techniques, such as direct preference optimization (DPO), has shown effective in enhancing the generalization abilities of large language model (LLM). However, in tasks involving video instruction-following, providing informative feedback, especially for open-ended conversations, remains a significant challenge. While previous studies have explored using large multimodal models (LMMs) as reward models for guiding preference modeling, their ability to accurately assess the quality of generated responses and their alignment with video content has not been conclusively demonstrated. This paper introduces a novel framework that utilizes detailed video captions as a proxy of video content, enabling language models to incorporate this information as supporting evidence for scoring video Question Answering (QA) predictions. Our approach demonstrates robust alignment with OpenAI GPT-4V model’s reward mechanism, which directly takes video frames as input. Furthermore, we show that applying our reward mechanism to DPO algorithm significantly improves model performance on open-ended video QA tasks.</abstract>
<identifier type="citekey">zhang-etal-2025-direct</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-long.30</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-long.30/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>694</start>
<end>717</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward
%A Zhang, Ruohong
%A Gui, Liangke
%A Sun, Zhiqing
%A Feng, Yihao
%A Xu, Keyang
%A Zhang, Yuanhan
%A Fu, Di
%A Li, Chunyuan
%A Hauptmann, Alexander G.
%A Bisk, Yonatan
%A Yang, Yiming
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-189-6
%F zhang-etal-2025-direct
%X Preference modeling techniques, such as direct preference optimization (DPO), has shown effective in enhancing the generalization abilities of large language model (LLM). However, in tasks involving video instruction-following, providing informative feedback, especially for open-ended conversations, remains a significant challenge. While previous studies have explored using large multimodal models (LMMs) as reward models for guiding preference modeling, their ability to accurately assess the quality of generated responses and their alignment with video content has not been conclusively demonstrated. This paper introduces a novel framework that utilizes detailed video captions as a proxy of video content, enabling language models to incorporate this information as supporting evidence for scoring video Question Answering (QA) predictions. Our approach demonstrates robust alignment with OpenAI GPT-4V model’s reward mechanism, which directly takes video frames as input. Furthermore, we show that applying our reward mechanism to DPO algorithm significantly improves model performance on open-ended video QA tasks.
%R 10.18653/v1/2025.naacl-long.30
%U https://aclanthology.org/2025.naacl-long.30/
%U https://doi.org/10.18653/v1/2025.naacl-long.30
%P 694-717
Markdown (Informal)
[Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward](https://aclanthology.org/2025.naacl-long.30/) (Zhang et al., NAACL 2025)
ACL
- Ruohong Zhang, Liangke Gui, Zhiqing Sun, Yihao Feng, Keyang Xu, Yuanhan Zhang, Di Fu, Chunyuan Li, Alexander G Hauptmann, Yonatan Bisk, and Yiming Yang. 2025. Direct Preference Optimization of Video Large Multimodal Models from Language Model Reward. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 694–717, Albuquerque, New Mexico. Association for Computational Linguistics.