Search papers, labs, and topics across Lattice.
2
0
4
Pretraining loss is a deceptive selection metric: at 30B MoE scale, downstream SFT performance is governed not by benchmark scores, but by the checkpoint's solution density under local weight perturbations.
Symbolic neural networks can provably recover the underlying PDE from measurement data, even with noisy inputs, opening the door to interpretable scientific discovery.
