Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Sapienza University of Rome
2
0
2
Traditional multimodal attention fails to capture joint interactions, but GRAMformer's volumetric approach reveals deeper dependencies and boosts performance.
Medical AI gets a boost: a new framework closes the "modality gap" in CLIP-based models, leading to better alignment between radiology images and clinical text.