Search papers, labs, and topics across Lattice.
1
0
3
0
By learning to focus on question-relevant image regions, Region-R1 boosts multi-modal re-ranking performance by up to 20%, showing that attending to the right visual cues is more important than seeing everything.
