Do Vision-Language Models (VLMs) actually "see" everything in a crowded room? š
Today at #CVPR2026, we are presenting VisualOverload, our work exploring the critical visual perception bottlenecks of VLMs in dense scenes.
š Today (Poster Session 6), 5:30 PM - 7:30 PM, Poster 431 (ExHall A)
