LV-XAttn: Distributed Cross-Attention for Long Visual Inputs in Multimodal Large Language Models.

Tzu-Tao Chang, Shivaram Venkataraman

Browse the full ICML paper archive.