Search papers, labs, and topics across Lattice.
2
0
3
0
Role-aware matching in image retrieval can boost performance by up to 23% without the need for task-specific training.
Hallucinations in LVLMs can be cut by over 43% without sacrificing grounded object coverage, thanks to a novel verifier-guided approach.
