@inproceedings{jeong-etal-2026-parentingllm,
title = "{P}arenting{LLM} : Pacing Reasoning Maturity for Bidirectional Test-Time Scaling under Token Budgets",
author = "Jeong, Seunghwan and
Lee, Sugi and
Park, WeiJin and
Yeom, Ikjun and
Woo, Honguk",
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.2244/",
pages = "51058--51071",
ISBN = "979-8-89176-459-0",
abstract = "Reasoning LLMs under fixed token budgets often commit before verification finishes, or spend the budget on deliberation that never reaches a final answer. Existing inference-time controllers regulate reasoning length, which does not indicate whether the current prefix is ready to be finalized. We introduce ParentingLLM, an inference-time controller that paces reasoning maturity, defined as the prefix{'}s readiness to commit given the remaining budget, against a user-specified target schedule, without fine-tuning the underlying LLM. The maturity estimate combines consumed budget with a forked forced-commit probe; the controller injects Wait when the estimate runs ahead of the schedule and Hurry when it lags. On MATH-500, GPQA-Diamond, AIME 24+25, and GSM-Hard with three Gemma-4 models and Qwen3.5-9B, ParentingLLM improves over prior inference-time baselines, with the largest gains on AIME 24+25 under a tight budget (-22.6 burst rate, +4.0 to +7.3 accuracy at B=1024)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jeong-etal-2026-parentingllm">
<titleInfo>
<title>ParentingLLM : Pacing Reasoning Maturity for Bidirectional Test-Time Scaling under Token Budgets</title>
</titleInfo>
<name type="personal">
<namePart type="given">Seunghwan</namePart>
<namePart type="family">Jeong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sugi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">WeiJin</namePart>
<namePart type="family">Park</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ikjun</namePart>
<namePart type="family">Yeom</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Honguk</namePart>
<namePart type="family">Woo</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>Reasoning LLMs under fixed token budgets often commit before verification finishes, or spend the budget on deliberation that never reaches a final answer. Existing inference-time controllers regulate reasoning length, which does not indicate whether the current prefix is ready to be finalized. We introduce ParentingLLM, an inference-time controller that paces reasoning maturity, defined as the prefix’s readiness to commit given the remaining budget, against a user-specified target schedule, without fine-tuning the underlying LLM. The maturity estimate combines consumed budget with a forked forced-commit probe; the controller injects Wait when the estimate runs ahead of the schedule and Hurry when it lags. On MATH-500, GPQA-Diamond, AIME 24+25, and GSM-Hard with three Gemma-4 models and Qwen3.5-9B, ParentingLLM improves over prior inference-time baselines, with the largest gains on AIME 24+25 under a tight budget (-22.6 burst rate, +4.0 to +7.3 accuracy at B=1024).</abstract>
<identifier type="citekey">jeong-etal-2026-parentingllm</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2244/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>51058</start>
<end>51071</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ParentingLLM : Pacing Reasoning Maturity for Bidirectional Test-Time Scaling under Token Budgets
%A Jeong, Seunghwan
%A Lee, Sugi
%A Park, WeiJin
%A Yeom, Ikjun
%A Woo, Honguk
%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 jeong-etal-2026-parentingllm
%X Reasoning LLMs under fixed token budgets often commit before verification finishes, or spend the budget on deliberation that never reaches a final answer. Existing inference-time controllers regulate reasoning length, which does not indicate whether the current prefix is ready to be finalized. We introduce ParentingLLM, an inference-time controller that paces reasoning maturity, defined as the prefix’s readiness to commit given the remaining budget, against a user-specified target schedule, without fine-tuning the underlying LLM. The maturity estimate combines consumed budget with a forked forced-commit probe; the controller injects Wait when the estimate runs ahead of the schedule and Hurry when it lags. On MATH-500, GPQA-Diamond, AIME 24+25, and GSM-Hard with three Gemma-4 models and Qwen3.5-9B, ParentingLLM improves over prior inference-time baselines, with the largest gains on AIME 24+25 under a tight budget (-22.6 burst rate, +4.0 to +7.3 accuracy at B=1024).
%U https://aclanthology.org/2026.emnlp-main.2244/
%P 51058-51071
Markdown (Informal)
[ParentingLLM : Pacing Reasoning Maturity for Bidirectional Test-Time Scaling under Token Budgets](https://aclanthology.org/2026.emnlp-main.2244/) (Jeong et al., EMNLP 2026)
ACL