Search papers, labs, and topics across Lattice.
1
0
3
Standard multimodal LLMs can perform surprisingly well on dense prediction tasks like segmentation and depth estimation, without needing any task-specific decoder modules.
