@inproceedings{ning-etal-2026-thoughts,
title = "Not All Thoughts are Generated Equal: Efficient {LLM} Reasoning via Synergizing-Oriented Multi-Turn Reinforcement Learning",
author = "Ning, Yansong and
Li, Wei and
Fang, Jun and
Tan, Naiqiang and
Liu, Hao",
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.916/",
pages = "21102--21126",
ISBN = "979-8-89176-459-0",
abstract = "Compressing long chain-of-thought (CoT) from large language models (LLMs) is an emerging strategy to improve reasoning efficiency. Despite its promising benefits, existing studies equally compress all thoughts within a long CoT, hindering more concise and effective reasoning. To this end, we first investigate the importance of different thoughts through automatic long CoT chunking and Monte Carlo rollouts. Building on this, we propose a theoretically bounded metric to jointly measure the effectiveness and efficiency of different thoughts. We then propose Long Short, an efficient reasoning framework that enables two LLMs to collaboratively solve problems: a long-thought LLM for effectively generating important thoughts, and a short-thought LLM for efficiently generating remaining thoughts. Specifically, we synthesize cold-start data to fine-tune LLMs for long-thought and short-thought reasoning styles, respectively. Furthermore, we propose synergizing-oriented multi-turn reinforcement learning, focusing on model self-evolution and collaboration between long-thought and short-thought LLMs. Experimental results show that our method outperforms seven efficient reasoning methods and enables Qwen2.5-7B to achieve comparable performance to DeepSeek-R1-Distill-Qwen-7B, while reducing token length by over 80{\%} across five reasoning benchmarks. Our code and data are available at https://github.com/usail-hkust/LongShort."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ning-etal-2026-thoughts">
<titleInfo>
<title>Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Synergizing-Oriented Multi-Turn Reinforcement Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yansong</namePart>
<namePart type="family">Ning</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Fang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naiqiang</namePart>
<namePart type="family">Tan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hao</namePart>
<namePart type="family">Liu</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>Compressing long chain-of-thought (CoT) from large language models (LLMs) is an emerging strategy to improve reasoning efficiency. Despite its promising benefits, existing studies equally compress all thoughts within a long CoT, hindering more concise and effective reasoning. To this end, we first investigate the importance of different thoughts through automatic long CoT chunking and Monte Carlo rollouts. Building on this, we propose a theoretically bounded metric to jointly measure the effectiveness and efficiency of different thoughts. We then propose Long Short, an efficient reasoning framework that enables two LLMs to collaboratively solve problems: a long-thought LLM for effectively generating important thoughts, and a short-thought LLM for efficiently generating remaining thoughts. Specifically, we synthesize cold-start data to fine-tune LLMs for long-thought and short-thought reasoning styles, respectively. Furthermore, we propose synergizing-oriented multi-turn reinforcement learning, focusing on model self-evolution and collaboration between long-thought and short-thought LLMs. Experimental results show that our method outperforms seven efficient reasoning methods and enables Qwen2.5-7B to achieve comparable performance to DeepSeek-R1-Distill-Qwen-7B, while reducing token length by over 80% across five reasoning benchmarks. Our code and data are available at https://github.com/usail-hkust/LongShort.</abstract>
<identifier type="citekey">ning-etal-2026-thoughts</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.916/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>21102</start>
<end>21126</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Synergizing-Oriented Multi-Turn Reinforcement Learning
%A Ning, Yansong
%A Li, Wei
%A Fang, Jun
%A Tan, Naiqiang
%A Liu, Hao
%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 ning-etal-2026-thoughts
%X Compressing long chain-of-thought (CoT) from large language models (LLMs) is an emerging strategy to improve reasoning efficiency. Despite its promising benefits, existing studies equally compress all thoughts within a long CoT, hindering more concise and effective reasoning. To this end, we first investigate the importance of different thoughts through automatic long CoT chunking and Monte Carlo rollouts. Building on this, we propose a theoretically bounded metric to jointly measure the effectiveness and efficiency of different thoughts. We then propose Long Short, an efficient reasoning framework that enables two LLMs to collaboratively solve problems: a long-thought LLM for effectively generating important thoughts, and a short-thought LLM for efficiently generating remaining thoughts. Specifically, we synthesize cold-start data to fine-tune LLMs for long-thought and short-thought reasoning styles, respectively. Furthermore, we propose synergizing-oriented multi-turn reinforcement learning, focusing on model self-evolution and collaboration between long-thought and short-thought LLMs. Experimental results show that our method outperforms seven efficient reasoning methods and enables Qwen2.5-7B to achieve comparable performance to DeepSeek-R1-Distill-Qwen-7B, while reducing token length by over 80% across five reasoning benchmarks. Our code and data are available at https://github.com/usail-hkust/LongShort.
%U https://aclanthology.org/2026.emnlp-main.916/
%P 21102-21126
Markdown (Informal)
[Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Synergizing-Oriented Multi-Turn Reinforcement Learning](https://aclanthology.org/2026.emnlp-main.916/) (Ning et al., EMNLP 2026)
ACL