@inproceedings{kaplan-etal-2026-follow,
title = "Follow the Flow: On Information Flow Across Textual Tokens in Text-to-Image Models",
author = "Kaplan, Guy and
Toker, Michael and
Reif, Yuval and
Belinkov, Yonatan and
Schwartz, Roy",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.1575/",
pages = "34139--34157",
ISBN = "979-8-89176-390-6",
abstract = "Text-to-image generation models suffer from alignment problems, where generated images fail to accurately capture the objects and relations in the text prompt. Prior work has focused on improving alignment by refining the diffusion process, ignoring the role of the text encoder, which guides the diffusion. In this work, we investigate how semantic information is distributed across token representations in text-to-image prompts, analyzing it at two levels: (1) in-item representation{---}whether individual tokens represent their lexical item (i.e., a word or expression conveying a single concept), and (2) cross-item interaction{---}whether information flows between tokens of different lexical items. We use patching techniques to uncover encoding patterns, and find that information is usually concentrated in only one or two of the item{'}s tokens; for example, in the item ``San Francisco{'}s Golden Gate Bridge'', the token ``Gate'' sufficiently captures the entire expression while the other tokens could effectively be discarded. Lexical items also tend to remain isolated; for instance, in the prompt ``a green dog'', the token ``dog'' encodes no visual information about ``green''. However, in some cases, items do influence each other{'}s representation, often leading to misinterpretations{---}e.g., in the prompt ``a pool by a table'', the token ``pool'' represents a ``pool table'' after contextualization. Our findings highlight the critical role of token-level encoding in image generation, and demonstrate that simple interventions at the encoding stage can substantially improve alignment and generation quality."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kaplan-etal-2026-follow">
<titleInfo>
<title>Follow the Flow: On Information Flow Across Textual Tokens in Text-to-Image Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Guy</namePart>
<namePart type="family">Kaplan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="family">Toker</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuval</namePart>
<namePart type="family">Reif</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yonatan</namePart>
<namePart type="family">Belinkov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roy</namePart>
<namePart type="family">Schwartz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-390-6</identifier>
</relatedItem>
<abstract>Text-to-image generation models suffer from alignment problems, where generated images fail to accurately capture the objects and relations in the text prompt. Prior work has focused on improving alignment by refining the diffusion process, ignoring the role of the text encoder, which guides the diffusion. In this work, we investigate how semantic information is distributed across token representations in text-to-image prompts, analyzing it at two levels: (1) in-item representation—whether individual tokens represent their lexical item (i.e., a word or expression conveying a single concept), and (2) cross-item interaction—whether information flows between tokens of different lexical items. We use patching techniques to uncover encoding patterns, and find that information is usually concentrated in only one or two of the item’s tokens; for example, in the item “San Francisco’s Golden Gate Bridge”, the token “Gate” sufficiently captures the entire expression while the other tokens could effectively be discarded. Lexical items also tend to remain isolated; for instance, in the prompt “a green dog”, the token “dog” encodes no visual information about “green”. However, in some cases, items do influence each other’s representation, often leading to misinterpretations—e.g., in the prompt “a pool by a table”, the token “pool” represents a “pool table” after contextualization. Our findings highlight the critical role of token-level encoding in image generation, and demonstrate that simple interventions at the encoding stage can substantially improve alignment and generation quality.</abstract>
<identifier type="citekey">kaplan-etal-2026-follow</identifier>
<location>
<url>https://aclanthology.org/2026.acl-long.1575/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>34139</start>
<end>34157</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Follow the Flow: On Information Flow Across Textual Tokens in Text-to-Image Models
%A Kaplan, Guy
%A Toker, Michael
%A Reif, Yuval
%A Belinkov, Yonatan
%A Schwartz, Roy
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F kaplan-etal-2026-follow
%X Text-to-image generation models suffer from alignment problems, where generated images fail to accurately capture the objects and relations in the text prompt. Prior work has focused on improving alignment by refining the diffusion process, ignoring the role of the text encoder, which guides the diffusion. In this work, we investigate how semantic information is distributed across token representations in text-to-image prompts, analyzing it at two levels: (1) in-item representation—whether individual tokens represent their lexical item (i.e., a word or expression conveying a single concept), and (2) cross-item interaction—whether information flows between tokens of different lexical items. We use patching techniques to uncover encoding patterns, and find that information is usually concentrated in only one or two of the item’s tokens; for example, in the item “San Francisco’s Golden Gate Bridge”, the token “Gate” sufficiently captures the entire expression while the other tokens could effectively be discarded. Lexical items also tend to remain isolated; for instance, in the prompt “a green dog”, the token “dog” encodes no visual information about “green”. However, in some cases, items do influence each other’s representation, often leading to misinterpretations—e.g., in the prompt “a pool by a table”, the token “pool” represents a “pool table” after contextualization. Our findings highlight the critical role of token-level encoding in image generation, and demonstrate that simple interventions at the encoding stage can substantially improve alignment and generation quality.
%U https://aclanthology.org/2026.acl-long.1575/
%P 34139-34157
Markdown (Informal)
[Follow the Flow: On Information Flow Across Textual Tokens in Text-to-Image Models](https://aclanthology.org/2026.acl-long.1575/) (Kaplan et al., ACL 2026)
ACL