SmellBench: Evaluating LLM Agents on Architectural Code Smell Repair

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Dinu, Ion George, Mihăescu, Marian Cristian, Rebedea, Traian
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911675164131328
author Dinu, Ion George
Mihăescu, Marian Cristian
Rebedea, Traian
author_facet Dinu, Ion George
Mihăescu, Marian Cristian
Rebedea, Traian
contents Architectural code smells erode software maintainability and are costly to repair manually, yet unlike localized bugs, they require cross-module reasoning about design intent that challenges both developers and automated tools. While large language model agents excel at bug fixing and code-level refactoring, their ability to repair architectural code smells remains unexplored. We present the first empirical evaluation of LLM agents on architectural code smell repair. We contribute SmellBench, a task orchestration framework that incorporates smell-type-specific optimized prompts and supports iterative multi-step execution, together with a scoring methodology that separately evaluates repair effectiveness, false positive identification, and net codebase impact. We evaluate 11 agent configurations from four model families (GPT, Claude, Gemini, Mistral) on 65 hard-severity architectural smells detected by PyExamine in the Python project scikit-learn, validated against expert judgments. Expert validation reveals that 63.1% of detected smells are false positives, while the best agent achieves a 47.7% resolution rate. Agents identify false positives with up to $κ= 0.94$ expert agreement, but repair aggressiveness and net codebase quality are inversely related: the most aggressive agent introduces 140 new smells. These findings expose a gap between current LLM capabilities in localized code transformations and the architectural understanding needed for cross-module refactoring. SmellBench provides reusable infrastructure for tracking progress on this underexplored dimension of automated software engineering. We release our code and data at https://doi.org/10.5281/zenodo.19247588.
format Preprint
id arxiv_https___arxiv_org_abs_2605_07001
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SmellBench: Evaluating LLM Agents on Architectural Code Smell Repair
Dinu, Ion George
Mihăescu, Marian Cristian
Rebedea, Traian
Software Engineering
Computation and Language
D.2.7; D.2.3; D.2.11; D.2.5; I.2.2
Architectural code smells erode software maintainability and are costly to repair manually, yet unlike localized bugs, they require cross-module reasoning about design intent that challenges both developers and automated tools. While large language model agents excel at bug fixing and code-level refactoring, their ability to repair architectural code smells remains unexplored. We present the first empirical evaluation of LLM agents on architectural code smell repair. We contribute SmellBench, a task orchestration framework that incorporates smell-type-specific optimized prompts and supports iterative multi-step execution, together with a scoring methodology that separately evaluates repair effectiveness, false positive identification, and net codebase impact. We evaluate 11 agent configurations from four model families (GPT, Claude, Gemini, Mistral) on 65 hard-severity architectural smells detected by PyExamine in the Python project scikit-learn, validated against expert judgments. Expert validation reveals that 63.1% of detected smells are false positives, while the best agent achieves a 47.7% resolution rate. Agents identify false positives with up to $κ= 0.94$ expert agreement, but repair aggressiveness and net codebase quality are inversely related: the most aggressive agent introduces 140 new smells. These findings expose a gap between current LLM capabilities in localized code transformations and the architectural understanding needed for cross-module refactoring. SmellBench provides reusable infrastructure for tracking progress on this underexplored dimension of automated software engineering. We release our code and data at https://doi.org/10.5281/zenodo.19247588.
title SmellBench: Evaluating LLM Agents on Architectural Code Smell Repair
topic Software Engineering
Computation and Language
D.2.7; D.2.3; D.2.11; D.2.5; I.2.2
url https://arxiv.org/abs/2605.07001