Latent Regularization in Generative Test Input Generation

I am happy to announce that our paper on the impact of latent space truncation for generating test inputs for DL classifiers has been accepted at the DeepTest2026, co-located with ICSE 2026. The study evaluates how different truncation strategies affect validity, diversity, and fault detection in StyleGAN-based test generation across MNIST, Fashion-MNIST, and CIFAR-10. This work complements our previous research on latent-based manipulation by systematically analyzing how latent variable regularization influences the quality of generated tests.

Abstract:

This study investigates the impact of regularization of latent spaces through truncation on the quality of generated test inputs for deep learning classifiers. We evaluate this effect using style-based GANs, a state-of-the-art generative approach, and assess quality along three dimensions: validity, diversity, and fault detection. We evaluate our approach on the boundary testing of deep learning image classifiers across three datasets, MNIST, Fashion MNIST, and CIFAR-10. We compare two truncation strategies: latent code mixing with binary search optimization and random latent truncation for generative exploration. Our experiments show that the latent code-mixing approach yields a higher fault detection rate than random truncation, while also improving both diversity and validity.