Search papers, labs, and topics across Lattice.
3
0
6
End-to-end driving gets a reliability boost: UniUncer's unified uncertainty framework slashes trajectory error by 7% and improves driving performance by 10.8% in challenging scenarios.
Uni-Flow achieves faster-than-real-time modeling of complex multiscale flows by decoupling temporal evolution and spatial refinement, opening doors for deployable surrogates in scientific machine learning.
Forget full attention: a hybrid sparse-linear attention model, MiniCPM-SALA, achieves 3.5x faster inference and supports 1M context length on a single GPU, all while maintaining comparable performance.
