@inproceedings{yun-etal-2026-distilling,
title = "Distilling Long-{C}o{T} Reasoning through Collaborative Step-wise Multi-Teacher Decoding",
author = "Yun, Taewon and
Shin, Jisu and
Choi, Jeonghwan and
Bang, Seunghwan and
Song, Hwanjun",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.1867/",
pages = "37452--37468",
ISBN = "979-8-89176-395-1",
abstract = "Distilling large reasoning models (LRMs) has become essential for making their Long-CoT reasoning capabilities practical, as full-scale inference remains computationally prohibitive. Existing curation-based approaches, which select complete reasoning traces post-hoc, overlook the collaborative potential of heterogeneous teachers and fail to adapt exploration dynamically, often leading to redundant sampling and missed opportunities for complementary reasoning. To address this, we introduce CoRD, a collaborative multi-teacher decoding framework that performs step-wise reasoning synthesis guided by predictive perplexity{--}based scoring and beam search. This approach enables heterogeneous LRMs to jointly construct coherent reasoning trajectories while maintaining diverse, high-potential hypotheses efficiently. Experiments show that CoRD generates higher-quality reasoning data and achieves student performance approaching teacher-level results, demonstrating that fine-grained collaboration among diverse LRMs yields structured, efficient, and robust reasoning distillation. The dataset and model are available at https://github.com/DISL-Lab/CoRD"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yun-etal-2026-distilling">
<titleInfo>
<title>Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding</title>
</titleInfo>
<name type="personal">
<namePart type="given">Taewon</namePart>
<namePart type="family">Yun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jisu</namePart>
<namePart type="family">Shin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeonghwan</namePart>
<namePart type="family">Choi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seunghwan</namePart>
<namePart type="family">Bang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hwanjun</namePart>
<namePart type="family">Song</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>Findings of the Association for Computational Linguistics: ACL 2026</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">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Distilling large reasoning models (LRMs) has become essential for making their Long-CoT reasoning capabilities practical, as full-scale inference remains computationally prohibitive. Existing curation-based approaches, which select complete reasoning traces post-hoc, overlook the collaborative potential of heterogeneous teachers and fail to adapt exploration dynamically, often leading to redundant sampling and missed opportunities for complementary reasoning. To address this, we introduce CoRD, a collaborative multi-teacher decoding framework that performs step-wise reasoning synthesis guided by predictive perplexity–based scoring and beam search. This approach enables heterogeneous LRMs to jointly construct coherent reasoning trajectories while maintaining diverse, high-potential hypotheses efficiently. Experiments show that CoRD generates higher-quality reasoning data and achieves student performance approaching teacher-level results, demonstrating that fine-grained collaboration among diverse LRMs yields structured, efficient, and robust reasoning distillation. The dataset and model are available at https://github.com/DISL-Lab/CoRD</abstract>
<identifier type="citekey">yun-etal-2026-distilling</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.1867/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>37452</start>
<end>37468</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding
%A Yun, Taewon
%A Shin, Jisu
%A Choi, Jeonghwan
%A Bang, Seunghwan
%A Song, Hwanjun
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F yun-etal-2026-distilling
%X Distilling large reasoning models (LRMs) has become essential for making their Long-CoT reasoning capabilities practical, as full-scale inference remains computationally prohibitive. Existing curation-based approaches, which select complete reasoning traces post-hoc, overlook the collaborative potential of heterogeneous teachers and fail to adapt exploration dynamically, often leading to redundant sampling and missed opportunities for complementary reasoning. To address this, we introduce CoRD, a collaborative multi-teacher decoding framework that performs step-wise reasoning synthesis guided by predictive perplexity–based scoring and beam search. This approach enables heterogeneous LRMs to jointly construct coherent reasoning trajectories while maintaining diverse, high-potential hypotheses efficiently. Experiments show that CoRD generates higher-quality reasoning data and achieves student performance approaching teacher-level results, demonstrating that fine-grained collaboration among diverse LRMs yields structured, efficient, and robust reasoning distillation. The dataset and model are available at https://github.com/DISL-Lab/CoRD
%U https://aclanthology.org/2026.findings-acl.1867/
%P 37452-37468
Markdown (Informal)
[Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding](https://aclanthology.org/2026.findings-acl.1867/) (Yun et al., Findings 2026)
ACL