Search papers, labs, and topics across Lattice.
2
4
2
2
Boundary jitter in chunked action heads can be effectively mitigated, leading to improved task success rates and temporal coherence in vision-language models.
Ditch the bounding boxes: DualDiff leverages Occupancy Ray-shape Sampling to generate driving scene videos with unprecedented fidelity, outperforming existing methods by a significant margin in both generation quality and downstream task performance.
