@inproceedings{gao-etal-2026-sgtr,
title = "{SGTR}: Learning Style-Generalizable Trajectory Reasoning for Zero-Shot Text Style Transfer",
author = "Gao, Yuan and
He, Yuwei and
Luo, Yashuo and
Hu, Junlin and
Yang, Jun",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.563/",
pages = "12935--12950",
ISBN = "979-8-89176-459-0",
abstract = "Text Style Transfer (TST) aims to rewrite text in a target style while preserving its semantics. However, existing methods rely on style-specific supervision and implicit generation, leading to poor generalization and semantic degradation given the inherent scarcity of parallel data. We propose Style-Generalizable Trajectory Reasoning (SGTR), a novel framework that formulates TST as the joint modeling of reasoning trajectories and outputs, decomposing the process into three interpretable stages: style understanding, deviation localization, and reflection refinement. To further enhance stylistic generalization, we introduce a latent contrastive learning objective, enabling the model to disentangle stylistic attributes from semantic content directly within its hidden representations. We distill these sophisticated analytical capabilities into a lightweight student model. Experiments demonstrate that SGTR consistently outperforms strong baselines in both style quality and semantic preservation, while exhibiting superior zero-shot generalization to unseen styles."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gao-etal-2026-sgtr">
<titleInfo>
<title>SGTR: Learning Style-Generalizable Trajectory Reasoning for Zero-Shot Text Style Transfer</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuan</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuwei</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yashuo</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junlin</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Yang</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Text Style Transfer (TST) aims to rewrite text in a target style while preserving its semantics. However, existing methods rely on style-specific supervision and implicit generation, leading to poor generalization and semantic degradation given the inherent scarcity of parallel data. We propose Style-Generalizable Trajectory Reasoning (SGTR), a novel framework that formulates TST as the joint modeling of reasoning trajectories and outputs, decomposing the process into three interpretable stages: style understanding, deviation localization, and reflection refinement. To further enhance stylistic generalization, we introduce a latent contrastive learning objective, enabling the model to disentangle stylistic attributes from semantic content directly within its hidden representations. We distill these sophisticated analytical capabilities into a lightweight student model. Experiments demonstrate that SGTR consistently outperforms strong baselines in both style quality and semantic preservation, while exhibiting superior zero-shot generalization to unseen styles.</abstract>
<identifier type="citekey">gao-etal-2026-sgtr</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.563/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>12935</start>
<end>12950</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SGTR: Learning Style-Generalizable Trajectory Reasoning for Zero-Shot Text Style Transfer
%A Gao, Yuan
%A He, Yuwei
%A Luo, Yashuo
%A Hu, Junlin
%A Yang, Jun
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F gao-etal-2026-sgtr
%X Text Style Transfer (TST) aims to rewrite text in a target style while preserving its semantics. However, existing methods rely on style-specific supervision and implicit generation, leading to poor generalization and semantic degradation given the inherent scarcity of parallel data. We propose Style-Generalizable Trajectory Reasoning (SGTR), a novel framework that formulates TST as the joint modeling of reasoning trajectories and outputs, decomposing the process into three interpretable stages: style understanding, deviation localization, and reflection refinement. To further enhance stylistic generalization, we introduce a latent contrastive learning objective, enabling the model to disentangle stylistic attributes from semantic content directly within its hidden representations. We distill these sophisticated analytical capabilities into a lightweight student model. Experiments demonstrate that SGTR consistently outperforms strong baselines in both style quality and semantic preservation, while exhibiting superior zero-shot generalization to unseen styles.
%U https://aclanthology.org/2026.emnlp-main.563/
%P 12935-12950
Markdown (Informal)
[SGTR: Learning Style-Generalizable Trajectory Reasoning for Zero-Shot Text Style Transfer](https://aclanthology.org/2026.emnlp-main.563/) (Gao et al., EMNLP 2026)
ACL