Danae Sánchez Villegas

VLMs can pinpoint frame-level anomalies but falter at detecting temporal inconsistencies, revealing a critical limitation in their reasoning capabilities.

VLMs are more easily swayed by misleading text than you think, and their impressive reasoning chains can mask, rather than reveal, this over-reliance on language.

Multimodal LLMs can spot subtle sexism as well as humans, but still miss the visual cues when multiple types of sexism occur together.