@inproceedings{fan-etal-2026-reliability,
title = "Reliability-Prioritized Fine-Grained Generation in Multimodal Large Language Models",
author = "Fan, Xiaomeng and
Wu, Wei and
Wu, Yuwei and
Gao, Zhi and
Shiyu, Luo and
Gao, Mingyang and
Zhao, Haoyu and
Diao, Zhenxin and
Ba, Yuxuan and
Feng, Lijia and
Jia, Yunde and
Harandi, Mehrtash",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.1814/",
pages = "38735--38778",
ISBN = "979-8-89176-462-0",
abstract = "Multimodal large language models (MLLMs) are increasingly expected to generate fine-grained descriptions of visual content. However, we observe and theoretically show that generating fine-grained responses poses a reliability challenge, i.e., fine-grained generation is more error-prone than coarse-grained generation. This phenomenon suggests that models should generate the finest description that remains reliable rather than simply produce more specific outputs. To investigate this problem, we develop GranFact, a granularity-aware benchmark consisting of expert-verified multi-object images with coarse-to-fine category annotations. Then, we design a hierarchy-aware evaluation algorithm, which assesses both whether model predictions are visually correct and how specific the correct predictions are. We also propose a reliability-prioritized preference optimization method based on Direct Preference Optimization (DPO), which penalizes unreliable fine-grained claims while rewarding reliable specificity. Experiments on GranFact show that our method improves fine-grained generation while preserving reliability. Code and data are available at https://github.com/WeiWu2025/GranFact."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="fan-etal-2026-reliability">
<titleInfo>
<title>Reliability-Prioritized Fine-Grained Generation in Multimodal Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiaomeng</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuwei</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhi</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Luo</namePart>
<namePart type="family">Shiyu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mingyang</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haoyu</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhenxin</namePart>
<namePart type="family">Diao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuxuan</namePart>
<namePart type="family">Ba</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lijia</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yunde</namePart>
<namePart type="family">Jia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mehrtash</namePart>
<namePart type="family">Harandi</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Multimodal large language models (MLLMs) are increasingly expected to generate fine-grained descriptions of visual content. However, we observe and theoretically show that generating fine-grained responses poses a reliability challenge, i.e., fine-grained generation is more error-prone than coarse-grained generation. This phenomenon suggests that models should generate the finest description that remains reliable rather than simply produce more specific outputs. To investigate this problem, we develop GranFact, a granularity-aware benchmark consisting of expert-verified multi-object images with coarse-to-fine category annotations. Then, we design a hierarchy-aware evaluation algorithm, which assesses both whether model predictions are visually correct and how specific the correct predictions are. We also propose a reliability-prioritized preference optimization method based on Direct Preference Optimization (DPO), which penalizes unreliable fine-grained claims while rewarding reliable specificity. Experiments on GranFact show that our method improves fine-grained generation while preserving reliability. Code and data are available at https://github.com/WeiWu2025/GranFact.</abstract>
<identifier type="citekey">fan-etal-2026-reliability</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1814/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>38735</start>
<end>38778</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Reliability-Prioritized Fine-Grained Generation in Multimodal Large Language Models
%A Fan, Xiaomeng
%A Wu, Wei
%A Wu, Yuwei
%A Gao, Zhi
%A Shiyu, Luo
%A Gao, Mingyang
%A Zhao, Haoyu
%A Diao, Zhenxin
%A Ba, Yuxuan
%A Feng, Lijia
%A Jia, Yunde
%A Harandi, Mehrtash
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F fan-etal-2026-reliability
%X Multimodal large language models (MLLMs) are increasingly expected to generate fine-grained descriptions of visual content. However, we observe and theoretically show that generating fine-grained responses poses a reliability challenge, i.e., fine-grained generation is more error-prone than coarse-grained generation. This phenomenon suggests that models should generate the finest description that remains reliable rather than simply produce more specific outputs. To investigate this problem, we develop GranFact, a granularity-aware benchmark consisting of expert-verified multi-object images with coarse-to-fine category annotations. Then, we design a hierarchy-aware evaluation algorithm, which assesses both whether model predictions are visually correct and how specific the correct predictions are. We also propose a reliability-prioritized preference optimization method based on Direct Preference Optimization (DPO), which penalizes unreliable fine-grained claims while rewarding reliable specificity. Experiments on GranFact show that our method improves fine-grained generation while preserving reliability. Code and data are available at https://github.com/WeiWu2025/GranFact.
%U https://aclanthology.org/2026.findings-emnlp.1814/
%P 38735-38778
Markdown (Informal)
[Reliability-Prioritized Fine-Grained Generation in Multimodal Large Language Models](https://aclanthology.org/2026.findings-emnlp.1814/) (Fan et al., Findings 2026)
ACL
- Xiaomeng Fan, Wei Wu, Yuwei Wu, Zhi Gao, Luo Shiyu, Mingyang Gao, Haoyu Zhao, Zhenxin Diao, Yuxuan Ba, Lijia Feng, Yunde Jia, and Mehrtash Harandi. 2026. Reliability-Prioritized Fine-Grained Generation in Multimodal Large Language Models. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 38735–38778, Budapest, Hungary. Association for Computational Linguistics.