Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
5
Models trained on VBVR-Pro not only excel in native visual reasoning tasks but also reveal critical insights into the effectiveness of different generative modalities.
A 1000x larger video reasoning dataset reveals early signs of emergent generalization, offering a new foundation for training and evaluating spatiotemporal AI.
