Search papers, labs, and topics across Lattice.
2
0
4
Generating geometrically consistent digital twins from video just got easier: Pantheon360 leverages 360掳 video diffusion and a 3D cache to achieve superior results.
VLMs can be significantly boosted on embodied tasks by mid-training on a carefully curated subset of VLM data that is highly aligned with the VLA domain, rivaling the performance of much larger models.
