Andreas Floros — accepted AI-conference papers

2026

On the Anisotropy of Score-Based Generative Models

Andreas Floros, Seyed-Mohsen Moosavi-Dezfooli, Pier Luigi Dragotti

ICML 2026poster

We investigate the role of network architecture in shaping the inductive biases of modern score-based generative models. To this end, we introduce the Score Anisotropy Directions (SADs), architecture-dependent directions that reveal how different networks preferentially capture data structure. Our a…

Cited by 0 Source ↗Scholar ↗

2025

Tracing the Roots: Leveraging Temporal Dynamics in Diffusion Trajectories for Origin Attribution

Andreas Floros, Seyed-Mohsen Moosavi-Dezfooli, Pier Luigi Dragotti

NeurIPS 2025poster

Diffusion models have transformed image synthesis through iterative denoising, by defining trajectories from noise to coherent data. While their capabilities are widely celebrated, a critical challenge remains unaddressed: ensuring responsible use by verifying whether an image originates from a mode…

Cited by 0 Source ↗Scholar ↗