Search papers, labs, and topics across Lattice.
1
0
2
A lightweight Q-value model can boost a 9B VLA's performance beyond that of a 27B model while reducing inference latency by 27%.
