@inproceedings{tamoyan-etal-2026-sherloc,
title = "{SHERLOC}: Structured Diagnostic Localization for Code Repair Agents",
author = "Tamoyan, Hovhannes and
Narenthiran, Sean and
Arakelyan, Erik and
Mezini, Mira and
Ginsburg, Boris",
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.711/",
pages = "16496--16524",
ISBN = "979-8-89176-459-0",
abstract = "LLM agents solve repository-level coding tasks through multi-turn tool use, but utilize half their budget on locating faults before editing. Dedicated localization frameworks have emerged, yet are still evaluated as file retrieval rather than actionable diagnosis, producing locations without the diagnostic context a repair agent needs. We introduce SHERLOC (*Structured Hypothesis-driven Exploration and Reasoning for Localization*), a training-free framework pairing a reasoning LLM with compact repository tools and self-recovery, without fine-tuning or multi-agent orchestration. SHERLOC reaches state-of-the-art localization across model scales: 84.33{\%} $\texttt{accuracy@}1$ on SWE-Bench Lite and 81.27{\%} $\texttt{recall@}1$ on SWE-Bench Verified; at $\sim30$B parameters, it matches or outperforms other agentic methods. Injecting our locations and diagnostic findings into repair agents yields an average $+5.95${~}pp resolve-rate gain from the best SHERLOC result per setting on SWE-Bench Verified. SHERLOC cuts localization and total tokens by 36.7{\%} and 23.1{\%} on average."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tamoyan-etal-2026-sherloc">
<titleInfo>
<title>SHERLOC: Structured Diagnostic Localization for Code Repair Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hovhannes</namePart>
<namePart type="family">Tamoyan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sean</namePart>
<namePart type="family">Narenthiran</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Erik</namePart>
<namePart type="family">Arakelyan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mira</namePart>
<namePart type="family">Mezini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Boris</namePart>
<namePart type="family">Ginsburg</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>LLM agents solve repository-level coding tasks through multi-turn tool use, but utilize half their budget on locating faults before editing. Dedicated localization frameworks have emerged, yet are still evaluated as file retrieval rather than actionable diagnosis, producing locations without the diagnostic context a repair agent needs. We introduce SHERLOC (*Structured Hypothesis-driven Exploration and Reasoning for Localization*), a training-free framework pairing a reasoning LLM with compact repository tools and self-recovery, without fine-tuning or multi-agent orchestration. SHERLOC reaches state-of-the-art localization across model scales: 84.33% accuracy@1 on SWE-Bench Lite and 81.27% recall@1 on SWE-Bench Verified; at \sim30B parameters, it matches or outperforms other agentic methods. Injecting our locations and diagnostic findings into repair agents yields an average +5.95 pp resolve-rate gain from the best SHERLOC result per setting on SWE-Bench Verified. SHERLOC cuts localization and total tokens by 36.7% and 23.1% on average.</abstract>
<identifier type="citekey">tamoyan-etal-2026-sherloc</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.711/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>16496</start>
<end>16524</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SHERLOC: Structured Diagnostic Localization for Code Repair Agents
%A Tamoyan, Hovhannes
%A Narenthiran, Sean
%A Arakelyan, Erik
%A Mezini, Mira
%A Ginsburg, Boris
%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 tamoyan-etal-2026-sherloc
%X LLM agents solve repository-level coding tasks through multi-turn tool use, but utilize half their budget on locating faults before editing. Dedicated localization frameworks have emerged, yet are still evaluated as file retrieval rather than actionable diagnosis, producing locations without the diagnostic context a repair agent needs. We introduce SHERLOC (*Structured Hypothesis-driven Exploration and Reasoning for Localization*), a training-free framework pairing a reasoning LLM with compact repository tools and self-recovery, without fine-tuning or multi-agent orchestration. SHERLOC reaches state-of-the-art localization across model scales: 84.33% accuracy@1 on SWE-Bench Lite and 81.27% recall@1 on SWE-Bench Verified; at \sim30B parameters, it matches or outperforms other agentic methods. Injecting our locations and diagnostic findings into repair agents yields an average +5.95 pp resolve-rate gain from the best SHERLOC result per setting on SWE-Bench Verified. SHERLOC cuts localization and total tokens by 36.7% and 23.1% on average.
%U https://aclanthology.org/2026.emnlp-main.711/
%P 16496-16524
Markdown (Informal)
[SHERLOC: Structured Diagnostic Localization for Code Repair Agents](https://aclanthology.org/2026.emnlp-main.711/) (Tamoyan et al., EMNLP 2026)
ACL
- Hovhannes Tamoyan, Sean Narenthiran, Erik Arakelyan, Mira Mezini, and Boris Ginsburg. 2026. SHERLOC: Structured Diagnostic Localization for Code Repair Agents. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 16496–16524, Budapest, Hungary. Association for Computational Linguistics.