@inproceedings{gee-etal-2025-code,
title = "Code-Optimise: Self-Generated Preference Data for Correctness and Efficiency",
author = "Gee, Leonidas and
Gritta, Milan and
Lampouras, Gerasimos and
Iacobacci, Ignacio",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Findings of the Association for Computational Linguistics: NAACL 2025",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-naacl.5/",
doi = "10.18653/v1/2025.findings-naacl.5",
pages = "79--94",
ISBN = "979-8-89176-195-7",
abstract = "Code Language Models have been trained togenerate accurate solutions, typically with noregard for runtime. On the other hand, previousworks that explored execution optimisationhave observed corresponding drops infunctional correctness. To that end, we introduceCode-Optimise, a framework that incorporatesboth correctness (passed, failed) andruntime (quick, slow) as learning signals viaself-generated preference data. Our frameworkis both lightweight and robust as it dynamicallyselects solutions to reduce overfitting whileavoiding a reliance on larger models for learningsignals. Code-Optimise achieves significantimprovements in pass@k while decreasingthe competitive baseline runtimes by anadditional 6{\%} for in-domain data and up to3{\%} for out-of-domain data. As a by-product,the average length of the generated solutionsis reduced by up to 48{\%} on MBPP and 23{\%}on HumanEval, resulting in faster and cheaperinference. The generated data and codebaseis open-sourced at https://github.com/huawei-noah/HEBO/tree/Code{\_}Optimise."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gee-etal-2025-code">
<titleInfo>
<title>Code-Optimise: Self-Generated Preference Data for Correctness and Efficiency</title>
</titleInfo>
<name type="personal">
<namePart type="given">Leonidas</namePart>
<namePart type="family">Gee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Milan</namePart>
<namePart type="family">Gritta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gerasimos</namePart>
<namePart type="family">Lampouras</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ignacio</namePart>
<namePart type="family">Iacobacci</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: NAACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-195-7</identifier>
</relatedItem>
<abstract>Code Language Models have been trained togenerate accurate solutions, typically with noregard for runtime. On the other hand, previousworks that explored execution optimisationhave observed corresponding drops infunctional correctness. To that end, we introduceCode-Optimise, a framework that incorporatesboth correctness (passed, failed) andruntime (quick, slow) as learning signals viaself-generated preference data. Our frameworkis both lightweight and robust as it dynamicallyselects solutions to reduce overfitting whileavoiding a reliance on larger models for learningsignals. Code-Optimise achieves significantimprovements in pass@k while decreasingthe competitive baseline runtimes by anadditional 6% for in-domain data and up to3% for out-of-domain data. As a by-product,the average length of the generated solutionsis reduced by up to 48% on MBPP and 23%on HumanEval, resulting in faster and cheaperinference. The generated data and codebaseis open-sourced at https://github.com/huawei-noah/HEBO/tree/Code_Optimise.</abstract>
<identifier type="citekey">gee-etal-2025-code</identifier>
<identifier type="doi">10.18653/v1/2025.findings-naacl.5</identifier>
<location>
<url>https://aclanthology.org/2025.findings-naacl.5/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>79</start>
<end>94</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Code-Optimise: Self-Generated Preference Data for Correctness and Efficiency
%A Gee, Leonidas
%A Gritta, Milan
%A Lampouras, Gerasimos
%A Iacobacci, Ignacio
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Findings of the Association for Computational Linguistics: NAACL 2025
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-195-7
%F gee-etal-2025-code
%X Code Language Models have been trained togenerate accurate solutions, typically with noregard for runtime. On the other hand, previousworks that explored execution optimisationhave observed corresponding drops infunctional correctness. To that end, we introduceCode-Optimise, a framework that incorporatesboth correctness (passed, failed) andruntime (quick, slow) as learning signals viaself-generated preference data. Our frameworkis both lightweight and robust as it dynamicallyselects solutions to reduce overfitting whileavoiding a reliance on larger models for learningsignals. Code-Optimise achieves significantimprovements in pass@k while decreasingthe competitive baseline runtimes by anadditional 6% for in-domain data and up to3% for out-of-domain data. As a by-product,the average length of the generated solutionsis reduced by up to 48% on MBPP and 23%on HumanEval, resulting in faster and cheaperinference. The generated data and codebaseis open-sourced at https://github.com/huawei-noah/HEBO/tree/Code_Optimise.
%R 10.18653/v1/2025.findings-naacl.5
%U https://aclanthology.org/2025.findings-naacl.5/
%U https://doi.org/10.18653/v1/2025.findings-naacl.5
%P 79-94
Markdown (Informal)
[Code-Optimise: Self-Generated Preference Data for Correctness and Efficiency](https://aclanthology.org/2025.findings-naacl.5/) (Gee et al., Findings 2025)
ACL