@inproceedings{zhou-etal-2026-paletteagent,
title = "{P}alette{A}gent: Self-Evolving Correction for Line-Art Colorization",
author = "Zhou, Binjia and
Wang, Guanghui and
Shan, Hui and
Chen, Weikai and
HU, Zeyu and
Yin, Yingda and
Zhang, Runze and
Yan, Kai and
Zhu, Lingting and
Wang, Xin and
Luo, Keyang",
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.289/",
pages = "6130--6158",
ISBN = "979-8-89176-462-0",
abstract = "Given a line-art sketch and user-specified color intent, controllable colorization must assign intended colors to fine-grained regions while preserving the underlying line structure. Existing methods have improved visual quality, but precise local adjustment remains difficult. In this work, we argue that the key limitation is the absence of a persistent correction state. Such a state records user intent, localized mismatches, grounded edit regions, and preservation constraints to guide targeted repairs. To address this, we introduce PaletteAgent, an agentic, self-evolving correction framework for controllable line-art colorization. PaletteAgent translates user intent into structured color prompts, employs judgment agents to diagnose mismatches, grounds problematic regions, preserves correct fields, and adaptively decides whether to refine globally, repair locally, or stop. This iterative process naturally defines the GRPO training objective for optimizing the agent{'}s correction policy. Empirical results show that PaletteAgent outperforms line-art colorization and general image-editing baselines. Further ablations confirm the role of maintaining a persistent correction state for controllable colorization."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhou-etal-2026-paletteagent">
<titleInfo>
<title>PaletteAgent: Self-Evolving Correction for Line-Art Colorization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Binjia</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guanghui</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hui</namePart>
<namePart type="family">Shan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weikai</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zeyu</namePart>
<namePart type="family">HU</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yingda</namePart>
<namePart type="family">Yin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Runze</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lingting</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Keyang</namePart>
<namePart type="family">Luo</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>Given a line-art sketch and user-specified color intent, controllable colorization must assign intended colors to fine-grained regions while preserving the underlying line structure. Existing methods have improved visual quality, but precise local adjustment remains difficult. In this work, we argue that the key limitation is the absence of a persistent correction state. Such a state records user intent, localized mismatches, grounded edit regions, and preservation constraints to guide targeted repairs. To address this, we introduce PaletteAgent, an agentic, self-evolving correction framework for controllable line-art colorization. PaletteAgent translates user intent into structured color prompts, employs judgment agents to diagnose mismatches, grounds problematic regions, preserves correct fields, and adaptively decides whether to refine globally, repair locally, or stop. This iterative process naturally defines the GRPO training objective for optimizing the agent’s correction policy. Empirical results show that PaletteAgent outperforms line-art colorization and general image-editing baselines. Further ablations confirm the role of maintaining a persistent correction state for controllable colorization.</abstract>
<identifier type="citekey">zhou-etal-2026-paletteagent</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.289/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>6130</start>
<end>6158</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PaletteAgent: Self-Evolving Correction for Line-Art Colorization
%A Zhou, Binjia
%A Wang, Guanghui
%A Shan, Hui
%A Chen, Weikai
%A HU, Zeyu
%A Yin, Yingda
%A Zhang, Runze
%A Yan, Kai
%A Zhu, Lingting
%A Wang, Xin
%A Luo, Keyang
%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 zhou-etal-2026-paletteagent
%X Given a line-art sketch and user-specified color intent, controllable colorization must assign intended colors to fine-grained regions while preserving the underlying line structure. Existing methods have improved visual quality, but precise local adjustment remains difficult. In this work, we argue that the key limitation is the absence of a persistent correction state. Such a state records user intent, localized mismatches, grounded edit regions, and preservation constraints to guide targeted repairs. To address this, we introduce PaletteAgent, an agentic, self-evolving correction framework for controllable line-art colorization. PaletteAgent translates user intent into structured color prompts, employs judgment agents to diagnose mismatches, grounds problematic regions, preserves correct fields, and adaptively decides whether to refine globally, repair locally, or stop. This iterative process naturally defines the GRPO training objective for optimizing the agent’s correction policy. Empirical results show that PaletteAgent outperforms line-art colorization and general image-editing baselines. Further ablations confirm the role of maintaining a persistent correction state for controllable colorization.
%U https://aclanthology.org/2026.findings-emnlp.289/
%P 6130-6158
Markdown (Informal)
[PaletteAgent: Self-Evolving Correction for Line-Art Colorization](https://aclanthology.org/2026.findings-emnlp.289/) (Zhou et al., Findings 2026)
ACL
- Binjia Zhou, Guanghui Wang, Hui Shan, Weikai Chen, Zeyu HU, Yingda Yin, Runze Zhang, Kai Yan, Lingting Zhu, Xin Wang, and Keyang Luo. 2026. PaletteAgent: Self-Evolving Correction for Line-Art Colorization. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 6130–6158, Budapest, Hungary. Association for Computational Linguistics.