2 papers
cs.CV2026
SCAM: A Real-World Typographic Robustness Evaluation for Multimodal Foundation Models
Justus Westerhoff, Erblina Purelku, Jakob Hackstein +4
Typographic attacks exploit the interplay between text and visual content in multimodal foundation models, causing misclassifications when misleading text is embedded within images…
cs.CV2025
Latent Diffusion U-Net Representations Contain Positional Embeddings and Anomalies
Jonas Loos, Lorenz Linhardt
Diffusion models have demonstrated remarkable capabilities in synthesizing realistic images, spurring interest in using their representations for various downstream tasks. To bette…