Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
2
0
Pre-trained robot policies can be adapted to new tasks with drastically improved out-of-distribution generalization by anchoring fine-tuning to retrieved pre-training representations.
Temporal action generation just got a major upgrade—TBD-VLA combines block diffusion with autoregressive decoding to boost inference speed and coherence.