Search papers, labs, and topics across Lattice.
2
0
4
3
Achieve high-fidelity, temporally coherent video editing without paired training data by combining sparse semantic control with dense motion and texture synthesis.
Current vision-language models stumble on subtle spine diagnoses, but a new dataset and benchmark expose these weaknesses and pave the way for clinically useful AI.
