Search papers, labs, and topics across Lattice.
Affiliation:
3
0
4
Topological protection fails dynamically in non-Hermitian systems when the skin effect overwhelms the Chern invariant, forcing supposedly robust boundary signals to disperse into the bulk.
Optimal learning rate and batch size in MoEs depend directly on the expert activation ratio rather than active or total parameter counts, unlocking predictable hyperparameter transfer even at 1/64 sparsity.
