@inproceedings{cao-etal-2026-mm,
title = "{MM}-{V}eri{R}ubric: Verifiable Evidence-Grounded Rubrics for Multimodal Reward Models",
author = "Cao, Zhiyong and
Wang, Bo and
Li, Jintao and
Liu, Shuxia and
Zhang, Kairan and
Wang, Aijia and
Wei, Yi-Liang and
Zhou, Sicheng",
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.2230/",
pages = "50766--50786",
ISBN = "979-8-89176-459-0",
abstract = "Reward modeling is essential for multimodal alignment, yet most existing reward models only provide outcome reward, which is sparse and fails to capture human preferences. Rubrics can provide structured and fine grained evaluation signals, but building effective rubrics for multimodal tasks is challenging. In this work, we propose **MM-VeriRubric**, a framework for scalable construction of training data for multimodal rubric generation and rubric-based reward models. To better use visual evidence in images, we introduce task-aware visual clue extraction, which makes rubric generation more grounded in task-relevant visual information. MM-VeriRubric then generates evidence-grounded rubrics and constructs rubric-aligned rejected responses with controlled error dimensions, enabling the quality of each rubric to be verified through consistency checking. The verified data are used to train both a rubric generator and a rubric-based reward model. Experiments on multimodal reward model benchmarks and downstream Best of N tasks show that MM-VeriRubric improves preference judgment (+16.4{\%} on VL-RewardBench) and provides more interpretable and diagnostic reward signals for multimodal alignment (+13.1{\%} on MMMU-Pro using Best of 5 selection). Code is available at: https://github.com/TraceLite-AI/MM-VeriRubric."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cao-etal-2026-mm">
<titleInfo>
<title>MM-VeriRubric: Verifiable Evidence-Grounded Rubrics for Multimodal Reward Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhiyong</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bo</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jintao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuxia</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kairan</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aijia</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi-Liang</namePart>
<namePart type="family">Wei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sicheng</namePart>
<namePart type="family">Zhou</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>Reward modeling is essential for multimodal alignment, yet most existing reward models only provide outcome reward, which is sparse and fails to capture human preferences. Rubrics can provide structured and fine grained evaluation signals, but building effective rubrics for multimodal tasks is challenging. In this work, we propose **MM-VeriRubric**, a framework for scalable construction of training data for multimodal rubric generation and rubric-based reward models. To better use visual evidence in images, we introduce task-aware visual clue extraction, which makes rubric generation more grounded in task-relevant visual information. MM-VeriRubric then generates evidence-grounded rubrics and constructs rubric-aligned rejected responses with controlled error dimensions, enabling the quality of each rubric to be verified through consistency checking. The verified data are used to train both a rubric generator and a rubric-based reward model. Experiments on multimodal reward model benchmarks and downstream Best of N tasks show that MM-VeriRubric improves preference judgment (+16.4% on VL-RewardBench) and provides more interpretable and diagnostic reward signals for multimodal alignment (+13.1% on MMMU-Pro using Best of 5 selection). Code is available at: https://github.com/TraceLite-AI/MM-VeriRubric.</abstract>
<identifier type="citekey">cao-etal-2026-mm</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2230/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>50766</start>
<end>50786</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MM-VeriRubric: Verifiable Evidence-Grounded Rubrics for Multimodal Reward Models
%A Cao, Zhiyong
%A Wang, Bo
%A Li, Jintao
%A Liu, Shuxia
%A Zhang, Kairan
%A Wang, Aijia
%A Wei, Yi-Liang
%A Zhou, Sicheng
%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 cao-etal-2026-mm
%X Reward modeling is essential for multimodal alignment, yet most existing reward models only provide outcome reward, which is sparse and fails to capture human preferences. Rubrics can provide structured and fine grained evaluation signals, but building effective rubrics for multimodal tasks is challenging. In this work, we propose **MM-VeriRubric**, a framework for scalable construction of training data for multimodal rubric generation and rubric-based reward models. To better use visual evidence in images, we introduce task-aware visual clue extraction, which makes rubric generation more grounded in task-relevant visual information. MM-VeriRubric then generates evidence-grounded rubrics and constructs rubric-aligned rejected responses with controlled error dimensions, enabling the quality of each rubric to be verified through consistency checking. The verified data are used to train both a rubric generator and a rubric-based reward model. Experiments on multimodal reward model benchmarks and downstream Best of N tasks show that MM-VeriRubric improves preference judgment (+16.4% on VL-RewardBench) and provides more interpretable and diagnostic reward signals for multimodal alignment (+13.1% on MMMU-Pro using Best of 5 selection). Code is available at: https://github.com/TraceLite-AI/MM-VeriRubric.
%U https://aclanthology.org/2026.emnlp-main.2230/
%P 50766-50786
Markdown (Informal)
[MM-VeriRubric: Verifiable Evidence-Grounded Rubrics for Multimodal Reward Models](https://aclanthology.org/2026.emnlp-main.2230/) (Cao et al., EMNLP 2026)
ACL
- Zhiyong Cao, Bo Wang, Jintao Li, Shuxia Liu, Kairan Zhang, Aijia Wang, Yi-Liang Wei, and Sicheng Zhou. 2026. MM-VeriRubric: Verifiable Evidence-Grounded Rubrics for Multimodal Reward Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 50766–50786, Budapest, Hungary. Association for Computational Linguistics.