@inproceedings{yan-etal-2026-reasonflow,
title = "{R}eason{F}low: Flow Matching that Thinks Step by Step",
author = "Yan, Bing and
Sun, Ge and
de Pablo, Juan J. and
Martiniani, Stefano and
Cho, Kyunghyun",
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.2268/",
pages = "48554--48571",
ISBN = "979-8-89176-462-0",
abstract = "Flow matching has shown strong promise as a continuous generative framework, but its standard formulation is not well-suited to reasoning problems. Conventional flow matching trains a model to transport noise directly toward a single final target. For tasks that require multi-step reasoning, this direct path is often too difficult: the final answer may be reachable only through a sequence of intermediate computations or transformations. Inspired by how autoregressive language models address this challenge through chain-of-thought supervision, we introduce ReasonFlow, a framework that brings the same scaffolding principle to flow matching. Instead of supervising the flow with only a final target, ReasonFlow constructs a sequence of CoT-induced anchor points derived from intermediate reasoning prefixes. During training, the clean target moves along a piecewise-linear path through these anchors before endpoint answer readout. This turns reasoning supervision from an output sequence into a trajectory-shaping signal for flow matching. Across arithmetic, chemistry, and logical-reasoning tasks, ReasonFlow substantially improves over no-CoT flow-matching baselines."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yan-etal-2026-reasonflow">
<titleInfo>
<title>ReasonFlow: Flow Matching that Thinks Step by Step</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bing</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ge</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Juan</namePart>
<namePart type="given">J</namePart>
<namePart type="family">de Pablo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Stefano</namePart>
<namePart type="family">Martiniani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kyunghyun</namePart>
<namePart type="family">Cho</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>Flow matching has shown strong promise as a continuous generative framework, but its standard formulation is not well-suited to reasoning problems. Conventional flow matching trains a model to transport noise directly toward a single final target. For tasks that require multi-step reasoning, this direct path is often too difficult: the final answer may be reachable only through a sequence of intermediate computations or transformations. Inspired by how autoregressive language models address this challenge through chain-of-thought supervision, we introduce ReasonFlow, a framework that brings the same scaffolding principle to flow matching. Instead of supervising the flow with only a final target, ReasonFlow constructs a sequence of CoT-induced anchor points derived from intermediate reasoning prefixes. During training, the clean target moves along a piecewise-linear path through these anchors before endpoint answer readout. This turns reasoning supervision from an output sequence into a trajectory-shaping signal for flow matching. Across arithmetic, chemistry, and logical-reasoning tasks, ReasonFlow substantially improves over no-CoT flow-matching baselines.</abstract>
<identifier type="citekey">yan-etal-2026-reasonflow</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2268/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>48554</start>
<end>48571</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ReasonFlow: Flow Matching that Thinks Step by Step
%A Yan, Bing
%A Sun, Ge
%A de Pablo, Juan J.
%A Martiniani, Stefano
%A Cho, Kyunghyun
%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 yan-etal-2026-reasonflow
%X Flow matching has shown strong promise as a continuous generative framework, but its standard formulation is not well-suited to reasoning problems. Conventional flow matching trains a model to transport noise directly toward a single final target. For tasks that require multi-step reasoning, this direct path is often too difficult: the final answer may be reachable only through a sequence of intermediate computations or transformations. Inspired by how autoregressive language models address this challenge through chain-of-thought supervision, we introduce ReasonFlow, a framework that brings the same scaffolding principle to flow matching. Instead of supervising the flow with only a final target, ReasonFlow constructs a sequence of CoT-induced anchor points derived from intermediate reasoning prefixes. During training, the clean target moves along a piecewise-linear path through these anchors before endpoint answer readout. This turns reasoning supervision from an output sequence into a trajectory-shaping signal for flow matching. Across arithmetic, chemistry, and logical-reasoning tasks, ReasonFlow substantially improves over no-CoT flow-matching baselines.
%U https://aclanthology.org/2026.findings-emnlp.2268/
%P 48554-48571
Markdown (Informal)
[ReasonFlow: Flow Matching that Thinks Step by Step](https://aclanthology.org/2026.findings-emnlp.2268/) (Yan et al., Findings 2026)
ACL
- Bing Yan, Ge Sun, Juan J. de Pablo, Stefano Martiniani, and Kyunghyun Cho. 2026. ReasonFlow: Flow Matching that Thinks Step by Step. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 48554–48571, Budapest, Hungary. Association for Computational Linguistics.