@inproceedings{das-etal-2025-yinyang,
title = "{Y}in{Y}ang-Align: A new Benchmark for Competing Objectives and Introducing Multi-Objective Preference based Text-to-Image Alignment",
author = "Das, Amitava and
Narsupalli, Yaswanth and
Singh, Gurpreet and
Jain, Vinija and
Sharma, Vasu and
Trivedy, Suranjana and
Chadha, Aman and
Sheth, Amit",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.1208/",
doi = "10.18653/v1/2025.findings-acl.1208",
pages = "23518--23598",
ISBN = "979-8-89176-256-5",
abstract = "Precise alignment in Text-to-Image (T2I) systems is crucial for generating visuals that reflect user intent while adhering to ethical and policy standards. Recent controversies, such as the Google Gemini-generated Pope image backlash, highlight the urgent need for robust alignment mechanisms. Building on alignment successes in Large Language Models (LLMs), this paper introduces YinYangAlign, a benchmarking framework designed to evaluate and optimize T2I systems across six inherently contradictory objectives. These objectives highlight core trade-offs, such as balancing faithfulness to prompts with artistic freedom and maintaining cultural sensitivity without compromising creativity. Alongside this benchmark, we propose the Contradictory Alignment Optimization (CAO) framework, an extension of Direct Preference Optimization (DPO), which employs multi-objective optimization techniques to address these competing goals. By leveraging per-axiom loss functions, synergy-driven global preferences, and innovative tools like the Synergy Jacobian, CAO achieves superior alignment across all objectives. Experimental results demonstrate significant improvements in fidelity, diversity, and ethical adherence, setting new benchmarks for the field. This work provides a scalable, effective approach to resolving alignment challenges in T2I systems while offering insights into broader AI alignment paradigms."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="das-etal-2025-yinyang">
<titleInfo>
<title>YinYang-Align: A new Benchmark for Competing Objectives and Introducing Multi-Objective Preference based Text-to-Image Alignment</title>
</titleInfo>
<name type="personal">
<namePart type="given">Amitava</namePart>
<namePart type="family">Das</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaswanth</namePart>
<namePart type="family">Narsupalli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gurpreet</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vinija</namePart>
<namePart type="family">Jain</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vasu</namePart>
<namePart type="family">Sharma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Suranjana</namePart>
<namePart type="family">Trivedy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aman</namePart>
<namePart type="family">Chadha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amit</namePart>
<namePart type="family">Sheth</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>Precise alignment in Text-to-Image (T2I) systems is crucial for generating visuals that reflect user intent while adhering to ethical and policy standards. Recent controversies, such as the Google Gemini-generated Pope image backlash, highlight the urgent need for robust alignment mechanisms. Building on alignment successes in Large Language Models (LLMs), this paper introduces YinYangAlign, a benchmarking framework designed to evaluate and optimize T2I systems across six inherently contradictory objectives. These objectives highlight core trade-offs, such as balancing faithfulness to prompts with artistic freedom and maintaining cultural sensitivity without compromising creativity. Alongside this benchmark, we propose the Contradictory Alignment Optimization (CAO) framework, an extension of Direct Preference Optimization (DPO), which employs multi-objective optimization techniques to address these competing goals. By leveraging per-axiom loss functions, synergy-driven global preferences, and innovative tools like the Synergy Jacobian, CAO achieves superior alignment across all objectives. Experimental results demonstrate significant improvements in fidelity, diversity, and ethical adherence, setting new benchmarks for the field. This work provides a scalable, effective approach to resolving alignment challenges in T2I systems while offering insights into broader AI alignment paradigms.</abstract>
<identifier type="citekey">das-etal-2025-yinyang</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.1208</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.1208/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>23518</start>
<end>23598</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T YinYang-Align: A new Benchmark for Competing Objectives and Introducing Multi-Objective Preference based Text-to-Image Alignment
%A Das, Amitava
%A Narsupalli, Yaswanth
%A Singh, Gurpreet
%A Jain, Vinija
%A Sharma, Vasu
%A Trivedy, Suranjana
%A Chadha, Aman
%A Sheth, Amit
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F das-etal-2025-yinyang
%X Precise alignment in Text-to-Image (T2I) systems is crucial for generating visuals that reflect user intent while adhering to ethical and policy standards. Recent controversies, such as the Google Gemini-generated Pope image backlash, highlight the urgent need for robust alignment mechanisms. Building on alignment successes in Large Language Models (LLMs), this paper introduces YinYangAlign, a benchmarking framework designed to evaluate and optimize T2I systems across six inherently contradictory objectives. These objectives highlight core trade-offs, such as balancing faithfulness to prompts with artistic freedom and maintaining cultural sensitivity without compromising creativity. Alongside this benchmark, we propose the Contradictory Alignment Optimization (CAO) framework, an extension of Direct Preference Optimization (DPO), which employs multi-objective optimization techniques to address these competing goals. By leveraging per-axiom loss functions, synergy-driven global preferences, and innovative tools like the Synergy Jacobian, CAO achieves superior alignment across all objectives. Experimental results demonstrate significant improvements in fidelity, diversity, and ethical adherence, setting new benchmarks for the field. This work provides a scalable, effective approach to resolving alignment challenges in T2I systems while offering insights into broader AI alignment paradigms.
%R 10.18653/v1/2025.findings-acl.1208
%U https://aclanthology.org/2025.findings-acl.1208/
%U https://doi.org/10.18653/v1/2025.findings-acl.1208
%P 23518-23598
Markdown (Informal)
[YinYang-Align: A new Benchmark for Competing Objectives and Introducing Multi-Objective Preference based Text-to-Image Alignment](https://aclanthology.org/2025.findings-acl.1208/) (Das et al., Findings 2025)
ACL
- Amitava Das, Yaswanth Narsupalli, Gurpreet Singh, Vinija Jain, Vasu Sharma, Suranjana Trivedy, Aman Chadha, and Amit Sheth. 2025. YinYang-Align: A new Benchmark for Competing Objectives and Introducing Multi-Objective Preference based Text-to-Image Alignment. In Findings of the Association for Computational Linguistics: ACL 2025, pages 23518–23598, Vienna, Austria. Association for Computational Linguistics.