Search papers, labs, and topics across Lattice.
1
0
2
6
Pretraining loss is a misleading metric, leading to inefficient compute use, while a simple scaling rule can unlock superior performance in pixel-wise Earth-observation models.
