2026
Logic Unseen: Revealing the Logical Blindspots of Vision-Language Models
AAAI 2026technical
Vision-Language Models (VLMs), exemplified by CLIP, have emerged as foundational for multimodal intelligence. However, their capacity for logical understanding remains significantly underexplored, resulting in critical **