@inproceedings{lin-etal-2026-towards,
title = "Towards Efficient Chain-of-Thought Reasoning via Adaptive-Budgeting based Policy Optimization",
author = "Lin, Wenhao and
Wu, Qiong and
Yang, Shuaixiang and
Zhou, Yiyi and
Li, Hui and
Chen, Kehai and
Sun, Xiaoshuai and
Ji, Rongrong",
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.1050/",
pages = "22223--22265",
ISBN = "979-8-89176-462-0",
abstract = "Chain-of-Thought (CoT) reasoning has become a key problem-solving capability for advanced large language models (LLMs), but the ``over-thinking'' issue still plagues LLMs. In this paper, we propose a novel learning scheme termed Adaptive-Budgeting based Policy Optimization (ABPO) to balance the performance and efficiency of CoT reasoning. Our ABPO defines the RL training as an adaptive curriculum learning process, where example pools are curated to categorize training examples into three types, namely the mastered, learning and hard ones, respectively. As the training progresses, ABPO will adaptively schedule the examples with proper length budgets, and the example pools will also be dynamically updated based on the model status, thereby achieving a good balance between efficiency and performance of CoTs. The experimental results not only show the substantial efficiency improvements by ABPO, e.g., reducing token length by 78.3{\%} while improving 2.0{\%} performance of DeepSeek-R1-Distill-Qwen-1.5B on average, but also show its obvious advantages over the compared methods, e.g., reducing 59.4{\%} length and increasing 8.3{\%} performance on average than HAPO, respectively. Our code is released at https://github.com/SenriAkane1/ABPO/"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lin-etal-2026-towards">
<titleInfo>
<title>Towards Efficient Chain-of-Thought Reasoning via Adaptive-Budgeting based Policy Optimization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wenhao</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qiong</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuaixiang</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiyi</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hui</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kehai</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaoshuai</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rongrong</namePart>
<namePart type="family">Ji</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>Chain-of-Thought (CoT) reasoning has become a key problem-solving capability for advanced large language models (LLMs), but the “over-thinking” issue still plagues LLMs. In this paper, we propose a novel learning scheme termed Adaptive-Budgeting based Policy Optimization (ABPO) to balance the performance and efficiency of CoT reasoning. Our ABPO defines the RL training as an adaptive curriculum learning process, where example pools are curated to categorize training examples into three types, namely the mastered, learning and hard ones, respectively. As the training progresses, ABPO will adaptively schedule the examples with proper length budgets, and the example pools will also be dynamically updated based on the model status, thereby achieving a good balance between efficiency and performance of CoTs. The experimental results not only show the substantial efficiency improvements by ABPO, e.g., reducing token length by 78.3% while improving 2.0% performance of DeepSeek-R1-Distill-Qwen-1.5B on average, but also show its obvious advantages over the compared methods, e.g., reducing 59.4% length and increasing 8.3% performance on average than HAPO, respectively. Our code is released at https://github.com/SenriAkane1/ABPO/</abstract>
<identifier type="citekey">lin-etal-2026-towards</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1050/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>22223</start>
<end>22265</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Towards Efficient Chain-of-Thought Reasoning via Adaptive-Budgeting based Policy Optimization
%A Lin, Wenhao
%A Wu, Qiong
%A Yang, Shuaixiang
%A Zhou, Yiyi
%A Li, Hui
%A Chen, Kehai
%A Sun, Xiaoshuai
%A Ji, Rongrong
%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 lin-etal-2026-towards
%X Chain-of-Thought (CoT) reasoning has become a key problem-solving capability for advanced large language models (LLMs), but the “over-thinking” issue still plagues LLMs. In this paper, we propose a novel learning scheme termed Adaptive-Budgeting based Policy Optimization (ABPO) to balance the performance and efficiency of CoT reasoning. Our ABPO defines the RL training as an adaptive curriculum learning process, where example pools are curated to categorize training examples into three types, namely the mastered, learning and hard ones, respectively. As the training progresses, ABPO will adaptively schedule the examples with proper length budgets, and the example pools will also be dynamically updated based on the model status, thereby achieving a good balance between efficiency and performance of CoTs. The experimental results not only show the substantial efficiency improvements by ABPO, e.g., reducing token length by 78.3% while improving 2.0% performance of DeepSeek-R1-Distill-Qwen-1.5B on average, but also show its obvious advantages over the compared methods, e.g., reducing 59.4% length and increasing 8.3% performance on average than HAPO, respectively. Our code is released at https://github.com/SenriAkane1/ABPO/
%U https://aclanthology.org/2026.findings-emnlp.1050/
%P 22223-22265
Markdown (Informal)
[Towards Efficient Chain-of-Thought Reasoning via Adaptive-Budgeting based Policy Optimization](https://aclanthology.org/2026.findings-emnlp.1050/) (Lin et al., Findings 2026)
ACL
- Wenhao Lin, Qiong Wu, Shuaixiang Yang, Yiyi Zhou, Hui Li, Kehai Chen, Xiaoshuai Sun, and Rongrong Ji. 2026. Towards Efficient Chain-of-Thought Reasoning via Adaptive-Budgeting based Policy Optimization. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 22223–22265, Budapest, Hungary. Association for Computational Linguistics.