Learning Robust Vision-Language Models from Natural Latent Spaces
NeurIPSSpotlight2025
TL;DR
Pre-trained vision-language models (VLMs) exhibit significant vulnerability to imperceptible adversarial perturbations…
Opening excerpt from the authors’ abstract. source
Read the paper
Topics
vision-language language model adversarial
← All NeurIPS 2025 Spotlight papers · Browse the whole archive