@inproceedings{gu-etal-2025-token,
title = "Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation",
author = "Gu, Jihao and
Wang, Yingyao and
Cao, Meng and
Bu, Pi and
Song, Jun and
Zheng, Bo and
He, Yancheng and
Li, Shilong",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-emnlp.1076/",
doi = "10.18653/v1/2025.findings-emnlp.1076",
pages = "19754--19767",
ISBN = "979-8-89176-335-7",
abstract = "Direct Preference Optimization (DPO) has been demonstrated to be highly effective in mitigating hallucinations in Large Vision Language Models (LVLMs) by aligning their outputs more closely with human preferences. Despite the recent progress, existing methods suffer from two drawbacks: 1) Lack of scalable token-level rewards; and 2) Neglect of visual-anchored tokens. To this end, we propose a novel Token Preference Optimization model with self-calibrated rewards (dubbed as TPO), which adaptively attends to visual correlated tokens without fine-grained annotations. Specifically, we introduce a token-level visual-anchored reward as the difference of the logistic distributions of generated tokens conditioned on the raw image and the corrupted one. In addition, to highlight the informative visual-anchored tokens, a visual-aware training objective is proposed to enhance more accurate token-level optimization. Extensive experimental results have manifested the state-of-the-art performance of the proposed TPO. For example, by building on top of LLaVA and Qwen, our TPO boosts the performance absolute improvement for hallucination benchmarks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gu-etal-2025-token">
<titleInfo>
<title>Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jihao</namePart>
<namePart type="family">Gu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yingyao</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Meng</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pi</namePart>
<namePart type="family">Bu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bo</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yancheng</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shilong</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-335-7</identifier>
</relatedItem>
<abstract>Direct Preference Optimization (DPO) has been demonstrated to be highly effective in mitigating hallucinations in Large Vision Language Models (LVLMs) by aligning their outputs more closely with human preferences. Despite the recent progress, existing methods suffer from two drawbacks: 1) Lack of scalable token-level rewards; and 2) Neglect of visual-anchored tokens. To this end, we propose a novel Token Preference Optimization model with self-calibrated rewards (dubbed as TPO), which adaptively attends to visual correlated tokens without fine-grained annotations. Specifically, we introduce a token-level visual-anchored reward as the difference of the logistic distributions of generated tokens conditioned on the raw image and the corrupted one. In addition, to highlight the informative visual-anchored tokens, a visual-aware training objective is proposed to enhance more accurate token-level optimization. Extensive experimental results have manifested the state-of-the-art performance of the proposed TPO. For example, by building on top of LLaVA and Qwen, our TPO boosts the performance absolute improvement for hallucination benchmarks.</abstract>
<identifier type="citekey">gu-etal-2025-token</identifier>
<identifier type="doi">10.18653/v1/2025.findings-emnlp.1076</identifier>
<location>
<url>https://aclanthology.org/2025.findings-emnlp.1076/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>19754</start>
<end>19767</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation
%A Gu, Jihao
%A Wang, Yingyao
%A Cao, Meng
%A Bu, Pi
%A Song, Jun
%A Zheng, Bo
%A He, Yancheng
%A Li, Shilong
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Findings of the Association for Computational Linguistics: EMNLP 2025
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-335-7
%F gu-etal-2025-token
%X Direct Preference Optimization (DPO) has been demonstrated to be highly effective in mitigating hallucinations in Large Vision Language Models (LVLMs) by aligning their outputs more closely with human preferences. Despite the recent progress, existing methods suffer from two drawbacks: 1) Lack of scalable token-level rewards; and 2) Neglect of visual-anchored tokens. To this end, we propose a novel Token Preference Optimization model with self-calibrated rewards (dubbed as TPO), which adaptively attends to visual correlated tokens without fine-grained annotations. Specifically, we introduce a token-level visual-anchored reward as the difference of the logistic distributions of generated tokens conditioned on the raw image and the corrupted one. In addition, to highlight the informative visual-anchored tokens, a visual-aware training objective is proposed to enhance more accurate token-level optimization. Extensive experimental results have manifested the state-of-the-art performance of the proposed TPO. For example, by building on top of LLaVA and Qwen, our TPO boosts the performance absolute improvement for hallucination benchmarks.
%R 10.18653/v1/2025.findings-emnlp.1076
%U https://aclanthology.org/2025.findings-emnlp.1076/
%U https://doi.org/10.18653/v1/2025.findings-emnlp.1076
%P 19754-19767
Markdown (Informal)
[Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation](https://aclanthology.org/2025.findings-emnlp.1076/) (Gu et al., Findings 2025)
ACL
- Jihao Gu, Yingyao Wang, Meng Cao, Pi Bu, Jun Song, Bo Zheng, Yancheng He, and Shilong Li. 2025. Token Preference Optimization with Self-Calibrated Visual-Anchored Rewards for Hallucination Mitigation. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 19754–19767, Suzhou, China. Association for Computational Linguistics.