Search papers, labs, and topics across Lattice.
Beihang University
3
0
5
WDL-OPD boosts MATH500 accuracy from 0.630 to 0.685, showcasing a powerful new approach to stabilizing on-policy distillation.
Trajectory anchoring bias in VLA models can be mitigated by transforming future trajectory decisions into verifiable selections, leading to more reliable reasoning in autonomous driving.
SpatialCLI enables VLMs to achieve a staggering 84.6% accuracy on spatial reasoning tasks, far surpassing existing models.
