@inproceedings{wen-etal-2026-budgetthinker,
title = "{B}udget{T}hinker: Empowering Budget-aware {LLM} Reasoning with Control Tokens",
author = "Wen, Hao and
Wu, Xinrui and
Sun, Yi and
Zhang, Feifei and
Chen, Liye and
Wang, Jie and
Su, Yifan and
Liu, Yunxin and
Zhang, Ya-Qin and
Li, Yuanchun",
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.1465/",
pages = "31182--31205",
ISBN = "979-8-89176-462-0",
abstract = "Scaling test-time computation improves the reasoning ability of large language models (LLMs), but the resulting latency and cost limit its use in latency- and cost-sensitive scenarios. We present BudgetThinker, a framework for reasoning-budget adherence that provides online control of chain-of-thought (CoT) length during decoding. Rather than relying on a one-shot prompt instruction, BudgetThinker periodically inserts specialized tokens that communicate the remaining reasoning budget. A two-stage training pipeline, consisting of supervised fine-tuning and curriculum-based reinforcement learning, teaches the model to interpret these signals and adapt its reasoning pace. Across the evaluated 1.5B and 7B text backbones, a 7B multimodal backbone, and a 1.5B code-reasoning model evaluated on HumanEval, BudgetThinker improves budget adherence and accuracy{--}efficiency trade-offs over competitive baselines. The controlled comparisons isolate the contribution of online ratio signaling from the shared training pipeline. Code is available at https://github.com/MobileLLM/BudgetThinker."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wen-etal-2026-budgetthinker">
<titleInfo>
<title>BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hao</namePart>
<namePart type="family">Wen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinrui</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yi</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Feifei</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liye</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jie</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yifan</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yunxin</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ya-Qin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuanchun</namePart>
<namePart type="family">Li</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>Scaling test-time computation improves the reasoning ability of large language models (LLMs), but the resulting latency and cost limit its use in latency- and cost-sensitive scenarios. We present BudgetThinker, a framework for reasoning-budget adherence that provides online control of chain-of-thought (CoT) length during decoding. Rather than relying on a one-shot prompt instruction, BudgetThinker periodically inserts specialized tokens that communicate the remaining reasoning budget. A two-stage training pipeline, consisting of supervised fine-tuning and curriculum-based reinforcement learning, teaches the model to interpret these signals and adapt its reasoning pace. Across the evaluated 1.5B and 7B text backbones, a 7B multimodal backbone, and a 1.5B code-reasoning model evaluated on HumanEval, BudgetThinker improves budget adherence and accuracy–efficiency trade-offs over competitive baselines. The controlled comparisons isolate the contribution of online ratio signaling from the shared training pipeline. Code is available at https://github.com/MobileLLM/BudgetThinker.</abstract>
<identifier type="citekey">wen-etal-2026-budgetthinker</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1465/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>31182</start>
<end>31205</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens
%A Wen, Hao
%A Wu, Xinrui
%A Sun, Yi
%A Zhang, Feifei
%A Chen, Liye
%A Wang, Jie
%A Su, Yifan
%A Liu, Yunxin
%A Zhang, Ya-Qin
%A Li, Yuanchun
%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 wen-etal-2026-budgetthinker
%X Scaling test-time computation improves the reasoning ability of large language models (LLMs), but the resulting latency and cost limit its use in latency- and cost-sensitive scenarios. We present BudgetThinker, a framework for reasoning-budget adherence that provides online control of chain-of-thought (CoT) length during decoding. Rather than relying on a one-shot prompt instruction, BudgetThinker periodically inserts specialized tokens that communicate the remaining reasoning budget. A two-stage training pipeline, consisting of supervised fine-tuning and curriculum-based reinforcement learning, teaches the model to interpret these signals and adapt its reasoning pace. Across the evaluated 1.5B and 7B text backbones, a 7B multimodal backbone, and a 1.5B code-reasoning model evaluated on HumanEval, BudgetThinker improves budget adherence and accuracy–efficiency trade-offs over competitive baselines. The controlled comparisons isolate the contribution of online ratio signaling from the shared training pipeline. Code is available at https://github.com/MobileLLM/BudgetThinker.
%U https://aclanthology.org/2026.findings-emnlp.1465/
%P 31182-31205
Markdown (Informal)
[BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens](https://aclanthology.org/2026.findings-emnlp.1465/) (Wen et al., Findings 2026)
ACL
- Hao Wen, Xinrui Wu, Yi Sun, Feifei Zhang, Liye Chen, Jie Wang, Yifan Su, Yunxin Liu, Ya-Qin Zhang, and Yuanchun Li. 2026. BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 31182–31205, Budapest, Hungary. Association for Computational Linguistics.