@inproceedings{li-etal-2026-writing,
title = "Writing As Reasoning: Interleaving Drafting and Deepening for Open-Ended Deep Research",
author = "Li, Yishan and
Chen, Wentong and
Yan, Yukun and
Li, Mingwei and
Liu, Kunpeng and
Cong, Xin and
Wang, Shuo and
Zhang, Zhong and
Lu, Yaxi and
Liu, Zhenghao and
Lin, Yankai and
Liu, Zhiyuan and
Sun, Maosong",
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.2336/",
pages = "50228--50249",
ISBN = "979-8-89176-462-0",
abstract = "Generating deep research reports requires large-scale information acquisition and the synthesis of insight-driven analyses, posing a major challenge for current language models. Existing methods largely follow a plan-then-write paradigm, whose performance strongly depends on the quality of the initial outline. We propose a Writing As Reasoning Policy (WARP) framework, which enables models to dynamically revise outlines during report writing. In this policy, the agent alternates between Evidence-Based Drafting and Reasoning-Driven Deepening phases, jointly supporting information acquisition, knowledge refinement, and outline updating. We further introduce a Multi-Stage Agentic Training{---}including cold start, atomic skill RL, and holistic pipeline RL{---}that enables small models to operate the WARP effectively. Experiments on DeepResearch Bench, DeepConsult, and DeepResearch Gym show that our approach allows small models to surpass leading closed-source systems, particularly with substantial improvements in Insight metric."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2026-writing">
<titleInfo>
<title>Writing As Reasoning: Interleaving Drafting and Deepening for Open-Ended Deep Research</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yishan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wentong</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yukun</namePart>
<namePart type="family">Yan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mingwei</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kunpeng</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xin</namePart>
<namePart type="family">Cong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuo</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhong</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaxi</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhenghao</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yankai</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiyuan</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maosong</namePart>
<namePart type="family">Sun</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>Generating deep research reports requires large-scale information acquisition and the synthesis of insight-driven analyses, posing a major challenge for current language models. Existing methods largely follow a plan-then-write paradigm, whose performance strongly depends on the quality of the initial outline. We propose a Writing As Reasoning Policy (WARP) framework, which enables models to dynamically revise outlines during report writing. In this policy, the agent alternates between Evidence-Based Drafting and Reasoning-Driven Deepening phases, jointly supporting information acquisition, knowledge refinement, and outline updating. We further introduce a Multi-Stage Agentic Training—including cold start, atomic skill RL, and holistic pipeline RL—that enables small models to operate the WARP effectively. Experiments on DeepResearch Bench, DeepConsult, and DeepResearch Gym show that our approach allows small models to surpass leading closed-source systems, particularly with substantial improvements in Insight metric.</abstract>
<identifier type="citekey">li-etal-2026-writing</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2336/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>50228</start>
<end>50249</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Writing As Reasoning: Interleaving Drafting and Deepening for Open-Ended Deep Research
%A Li, Yishan
%A Chen, Wentong
%A Yan, Yukun
%A Li, Mingwei
%A Liu, Kunpeng
%A Cong, Xin
%A Wang, Shuo
%A Zhang, Zhong
%A Lu, Yaxi
%A Liu, Zhenghao
%A Lin, Yankai
%A Liu, Zhiyuan
%A Sun, Maosong
%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 li-etal-2026-writing
%X Generating deep research reports requires large-scale information acquisition and the synthesis of insight-driven analyses, posing a major challenge for current language models. Existing methods largely follow a plan-then-write paradigm, whose performance strongly depends on the quality of the initial outline. We propose a Writing As Reasoning Policy (WARP) framework, which enables models to dynamically revise outlines during report writing. In this policy, the agent alternates between Evidence-Based Drafting and Reasoning-Driven Deepening phases, jointly supporting information acquisition, knowledge refinement, and outline updating. We further introduce a Multi-Stage Agentic Training—including cold start, atomic skill RL, and holistic pipeline RL—that enables small models to operate the WARP effectively. Experiments on DeepResearch Bench, DeepConsult, and DeepResearch Gym show that our approach allows small models to surpass leading closed-source systems, particularly with substantial improvements in Insight metric.
%U https://aclanthology.org/2026.findings-emnlp.2336/
%P 50228-50249
Markdown (Informal)
[Writing As Reasoning: Interleaving Drafting and Deepening for Open-Ended Deep Research](https://aclanthology.org/2026.findings-emnlp.2336/) (Li et al., Findings 2026)
ACL
- Yishan Li, Wentong Chen, Yukun Yan, Mingwei Li, Kunpeng Liu, Xin Cong, Shuo Wang, Zhong Zhang, Yaxi Lu, Zhenghao Liu, Yankai Lin, Zhiyuan Liu, and Maosong Sun. 2026. Writing As Reasoning: Interleaving Drafting and Deepening for Open-Ended Deep Research. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 50228–50249, Budapest, Hungary. Association for Computational Linguistics.