@inproceedings{edwards-etal-2026-rexbench,
title = "{RE}x{B}ench: Can coding agents autonomously implement {AI} research extensions?",
author = "Edwards, Nicholas and
Lee, Yukyung and
Mao, Yujun Audrey and
Qin, Yulu and
Schuster, Sebastian and
Kim, Najoung",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.745/",
doi = "10.18653/v1/2026.acl-long.745",
pages = "16380--16417",
ISBN = "979-8-89176-390-6",
abstract = "Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of realistic extensions of 12 research papers that aim to investigate \textit{novel} research hypotheses. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate 12 LLM agents implemented using two different frameworks: aider and OpenHands. We find that all agents fail to autonomously implement the majority of the extensions, with the best agent at around 33{\%} success rate. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 44{\%}. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="edwards-etal-2026-rexbench">
<titleInfo>
<title>RExBench: Can coding agents autonomously implement AI research extensions?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nicholas</namePart>
<namePart type="family">Edwards</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yukyung</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yujun</namePart>
<namePart type="given">Audrey</namePart>
<namePart type="family">Mao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yulu</namePart>
<namePart type="family">Qin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Schuster</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Najoung</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-390-6</identifier>
</relatedItem>
<abstract>Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of realistic extensions of 12 research papers that aim to investigate novel research hypotheses. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate 12 LLM agents implemented using two different frameworks: aider and OpenHands. We find that all agents fail to autonomously implement the majority of the extensions, with the best agent at around 33% success rate. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 44%. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance.</abstract>
<identifier type="citekey">edwards-etal-2026-rexbench</identifier>
<identifier type="doi">10.18653/v1/2026.acl-long.745</identifier>
<location>
<url>https://aclanthology.org/2026.acl-long.745/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>16380</start>
<end>16417</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T RExBench: Can coding agents autonomously implement AI research extensions?
%A Edwards, Nicholas
%A Lee, Yukyung
%A Mao, Yujun Audrey
%A Qin, Yulu
%A Schuster, Sebastian
%A Kim, Najoung
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F edwards-etal-2026-rexbench
%X Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of realistic extensions of 12 research papers that aim to investigate novel research hypotheses. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate 12 LLM agents implemented using two different frameworks: aider and OpenHands. We find that all agents fail to autonomously implement the majority of the extensions, with the best agent at around 33% success rate. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 44%. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance.
%R 10.18653/v1/2026.acl-long.745
%U https://aclanthology.org/2026.acl-long.745/
%U https://doi.org/10.18653/v1/2026.acl-long.745
%P 16380-16417
Markdown (Informal)
[RExBench: Can coding agents autonomously implement AI research extensions?](https://aclanthology.org/2026.acl-long.745/) (Edwards et al., ACL 2026)
ACL
- Nicholas Edwards, Yukyung Lee, Yujun Audrey Mao, Yulu Qin, Sebastian Schuster, and Najoung Kim. 2026. RExBench: Can coding agents autonomously implement AI research extensions?. In Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 16380–16417, San Diego, California, United States. Association for Computational Linguistics.