Latent Regularization in Generative Test Input Generation Conference: 7th International Workshop on Deep Learning for Testing and Testing for Deep Learning (DeepTest ’26); April 12–18, 2026; Rio de Janeiro, Brazil 7th International Workshop on Deep Learning for Testing and Testing for Deep Learning
Giorgi Merabishvili, Oliver Weißl, Andrea Stocco · 2026
Read Latent Regularization in Generative Test Input Generation Conference: 7th International Workshop on Deep Learning for Testing and Testing for Deep Learning (DeepTest ’26); April 12–18, 2026; Rio de Janeiro, Brazil 7th International Workshop on Deep Learning for Testing and Testing for Deep Learning by Giorgi Merabishvili, Oliver Weißl, Andrea Stocco (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-15552.txt.
Description
Merabishvili and colleagues test whether trimming the latent space of a generative model yields better test inputs for image classifiers. They work with style-based GANs and judge generated inputs on three fronts: whether they are valid, how diverse they are, and how often they expose faults. The test ground is boundary testing of classifiers on MNIST, Fashion MNIST, and CIFAR-10.
Two trimming routes are compared: mixing latent codes with binary search, and random latent truncation for open exploration. Code mixing wins, giving a higher fault detection rate while also improving diversity and validity. The paper offers testers a practical rule: shape the latent space with care, since how inputs are sampled changes what bugs surface.
Catalog record
| CREATOR | Giorgi Merabishvili, Oliver Weißl, Andrea Stocco |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-15552