Search papers, labs, and topics across Lattice.
2
0
4
Give your memory-less VLA policy a brain: TempoFit retrofits temporal context by cleverly reusing existing attention keys and values, boosting long-horizon task success without retraining or adding latency.
Achieve a 37% success rate boost in precision insertion by intelligently fusing human corrections with learned policies, triggered only when force discrepancies signal impending failure.
