@inproceedings{chen-etal-2026-logitscoder,
title = "{L}ogits{C}oder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation",
author = "Chen, Jizheng and
Zhang, Weiming and
Dai, Xinyi and
Fu, Lingyue and
Liu, Weiwen and
Du, Kounianhua and
Li, Qingyao and
Yu, Yong and
Zhang, Weinan",
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.1527/",
pages = "32495--32510",
ISBN = "979-8-89176-462-0",
abstract = "Code generation requires executable reasoning, making test-time chain-of-thought (CoT) search an attractive strategy for improving solution quality. However, existing CoT scaling methods often face a tension between underthinking, where shallow reasoning misses important algorithmic cases, and overthinking, where verbose or redundant reasoning consumes substantial tokens without improving correctness. We propose LogitsCoder, a lightweight framework that revisits CoT search from the perspective of logits-level control. LogitsCoder improves efficient reasoning search through two complementary mechanisms: Logits Preference Decoding adaptively calibrates local token expansion using public-test feedback, while Logits Rank Based Path Selection explores and selects compact high-confidence reasoning trajectories based on logit-level stability. Together with Thoughts Aggregation, LogitsCoder refines reasoning paths without relying on costly MCTS-style tree expansion. Extensive experiments demonstrate that LogitsCoder produces more efficient and higher-quality reasoning chains, leading to superior code generation performance compared to baseline methods."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-logitscoder">
<titleInfo>
<title>LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jizheng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weiming</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinyi</namePart>
<namePart type="family">Dai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lingyue</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weiwen</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kounianhua</namePart>
<namePart type="family">Du</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qingyao</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yong</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weinan</namePart>
<namePart type="family">Zhang</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>Code generation requires executable reasoning, making test-time chain-of-thought (CoT) search an attractive strategy for improving solution quality. However, existing CoT scaling methods often face a tension between underthinking, where shallow reasoning misses important algorithmic cases, and overthinking, where verbose or redundant reasoning consumes substantial tokens without improving correctness. We propose LogitsCoder, a lightweight framework that revisits CoT search from the perspective of logits-level control. LogitsCoder improves efficient reasoning search through two complementary mechanisms: Logits Preference Decoding adaptively calibrates local token expansion using public-test feedback, while Logits Rank Based Path Selection explores and selects compact high-confidence reasoning trajectories based on logit-level stability. Together with Thoughts Aggregation, LogitsCoder refines reasoning paths without relying on costly MCTS-style tree expansion. Extensive experiments demonstrate that LogitsCoder produces more efficient and higher-quality reasoning chains, leading to superior code generation performance compared to baseline methods.</abstract>
<identifier type="citekey">chen-etal-2026-logitscoder</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1527/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>32495</start>
<end>32510</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation
%A Chen, Jizheng
%A Zhang, Weiming
%A Dai, Xinyi
%A Fu, Lingyue
%A Liu, Weiwen
%A Du, Kounianhua
%A Li, Qingyao
%A Yu, Yong
%A Zhang, Weinan
%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 chen-etal-2026-logitscoder
%X Code generation requires executable reasoning, making test-time chain-of-thought (CoT) search an attractive strategy for improving solution quality. However, existing CoT scaling methods often face a tension between underthinking, where shallow reasoning misses important algorithmic cases, and overthinking, where verbose or redundant reasoning consumes substantial tokens without improving correctness. We propose LogitsCoder, a lightweight framework that revisits CoT search from the perspective of logits-level control. LogitsCoder improves efficient reasoning search through two complementary mechanisms: Logits Preference Decoding adaptively calibrates local token expansion using public-test feedback, while Logits Rank Based Path Selection explores and selects compact high-confidence reasoning trajectories based on logit-level stability. Together with Thoughts Aggregation, LogitsCoder refines reasoning paths without relying on costly MCTS-style tree expansion. Extensive experiments demonstrate that LogitsCoder produces more efficient and higher-quality reasoning chains, leading to superior code generation performance compared to baseline methods.
%U https://aclanthology.org/2026.findings-emnlp.1527/
%P 32495-32510
Markdown (Informal)
[LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation](https://aclanthology.org/2026.findings-emnlp.1527/) (Chen et al., Findings 2026)
ACL
- Jizheng Chen, Weiming Zhang, Xinyi Dai, Lingyue Fu, Weiwen Liu, Kounianhua Du, Qingyao Li, Yong Yu, and Weinan Zhang. 2026. LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 32495–32510, Budapest, Hungary. Association for Computational Linguistics.