Towards Automated Detection of Latent Software Robustness Failures Using Large Language Models
Year: 2026
Type: conference-paper
Abstract: Robustness failures occur when a system deviates from its specification under invalid inputs or stressful environmental conditions. Automated detection of these failures poses a substantial challenge, especially in the face of incomplete specifications and latent failure modes that do not produce explicit indicators. In this work, we propose the development of an approach for the automated detection of latent robustness failures in software systems by leveraging the reasoning capabilities of Large Language Models (LLMs). In particular, we define a pipeline that integrates multiple stages of robustness assessment, including enhancement of specifications, testing, and analysis of system behaviors, enabling a holistic and scalable approach.
Authors: José Flauzino, Marco Vieira, Elias P. Duarte
Language: English
Subfield: Software
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204771027
