@inproceedings{wang-etal-2026-bora,
title = "{BORA}-{BUDGET}: Learning When to Think Under a Token Budget",
author = "Wang, Zilong and
Pan, Kang and
Zhang, Xiao and
Meng, Erli and
Li, Shuai",
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.408/",
pages = "9380--9400",
ISBN = "979-8-89176-459-0",
abstract = "Hybrid-thinking language models expose both cheap non-thinking and expensive thinking modes, creating a deployment problem: deciding when additional reasoning is worth its token cost and override risk. We formulate this problem as budgeted opportunity allocation, where a controller first obtains a cheap seed answer, estimates the marginal value and risk of paid thinking, and preserves the seed unless an adoption gate accepts the thinking result. We propose \textsc{Bora-Budget}, a budget-aware inference framework for frozen LLMs, and instantiate it as \textsc{Bora-Switch-Calibrated}. The controller uses seed-state features to predict helpfulness, harmfulness, and thinking cost, then allocates thinking through either frontier operating points or an AutoThreshold policy calibrated to a total token budget. Across LLM scales on math dataset, the method substantially improves over random allocation and approaches uniform-thinking accuracy under lower token budgets. Transfer results on multiple-choice, code, and hard-QA tasks show promising but uneven generalization: trace length remains a strong baseline, harmful overrides are reduced but not eliminated, and learning how much to think remains open. These results position \textsc{Bora-Budget} as a seed-preserving control layer for budgeted inference rather than a replacement for stronger always-thinking baselines."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-bora">
<titleInfo>
<title>BORA-BUDGET: Learning When to Think Under a Token Budget</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zilong</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kang</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiao</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Erli</namePart>
<namePart type="family">Meng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuai</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>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>Hybrid-thinking language models expose both cheap non-thinking and expensive thinking modes, creating a deployment problem: deciding when additional reasoning is worth its token cost and override risk. We formulate this problem as budgeted opportunity allocation, where a controller first obtains a cheap seed answer, estimates the marginal value and risk of paid thinking, and preserves the seed unless an adoption gate accepts the thinking result. We propose Bora-Budget, a budget-aware inference framework for frozen LLMs, and instantiate it as Bora-Switch-Calibrated. The controller uses seed-state features to predict helpfulness, harmfulness, and thinking cost, then allocates thinking through either frontier operating points or an AutoThreshold policy calibrated to a total token budget. Across LLM scales on math dataset, the method substantially improves over random allocation and approaches uniform-thinking accuracy under lower token budgets. Transfer results on multiple-choice, code, and hard-QA tasks show promising but uneven generalization: trace length remains a strong baseline, harmful overrides are reduced but not eliminated, and learning how much to think remains open. These results position Bora-Budget as a seed-preserving control layer for budgeted inference rather than a replacement for stronger always-thinking baselines.</abstract>
<identifier type="citekey">wang-etal-2026-bora</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.408/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>9380</start>
<end>9400</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T BORA-BUDGET: Learning When to Think Under a Token Budget
%A Wang, Zilong
%A Pan, Kang
%A Zhang, Xiao
%A Meng, Erli
%A Li, Shuai
%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 wang-etal-2026-bora
%X Hybrid-thinking language models expose both cheap non-thinking and expensive thinking modes, creating a deployment problem: deciding when additional reasoning is worth its token cost and override risk. We formulate this problem as budgeted opportunity allocation, where a controller first obtains a cheap seed answer, estimates the marginal value and risk of paid thinking, and preserves the seed unless an adoption gate accepts the thinking result. We propose Bora-Budget, a budget-aware inference framework for frozen LLMs, and instantiate it as Bora-Switch-Calibrated. The controller uses seed-state features to predict helpfulness, harmfulness, and thinking cost, then allocates thinking through either frontier operating points or an AutoThreshold policy calibrated to a total token budget. Across LLM scales on math dataset, the method substantially improves over random allocation and approaches uniform-thinking accuracy under lower token budgets. Transfer results on multiple-choice, code, and hard-QA tasks show promising but uneven generalization: trace length remains a strong baseline, harmful overrides are reduced but not eliminated, and learning how much to think remains open. These results position Bora-Budget as a seed-preserving control layer for budgeted inference rather than a replacement for stronger always-thinking baselines.
%U https://aclanthology.org/2026.emnlp-main.408/
%P 9380-9400
Markdown (Informal)
[BORA-BUDGET: Learning When to Think Under a Token Budget](https://aclanthology.org/2026.emnlp-main.408/) (Wang et al., EMNLP 2026)
ACL
- Zilong Wang, Kang Pan, Xiao Zhang, Erli Meng, and Shuai Li. 2026. BORA-BUDGET: Learning When to Think Under a Token Budget. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 9380–9400, Budapest, Hungary. Association for Computational Linguistics.