Search papers, labs, and topics across Lattice.
1
0
2
Captions can be made 48% more complete and hallucination reduced by 45% without retraining models, thanks to a novel prominence-aware rectification approach.
