Search papers, labs, and topics across Lattice.
Affiliation:
3
0
3
Oversmoothing in graph neural networks can be effectively tackled by a new index-theoretic criterion that reveals the true discriminative power of sheaf configurations.
CIPE transforms how Transformers encode positional information in non-Euclidean graphs, achieving a remarkable 35.5% performance boost over traditional methods.
By explicitly modeling periodicity and many-body interactions, Periodic-TDL doesn't just predict polymer properties better, it actually uncovers the underlying physical effects of functional group modifications.
