Search papers, labs, and topics across Lattice.
2
0
3
HeteroMosaic achieves up to 2.05X faster inference and 45.3% lower energy consumption by intelligently coordinating heterogeneous resources in edge AI systems.
TileFuse achieves up to 281% performance gains for GEMV operations, making AMD NPUs a game-changer for efficient LLM inference on edge devices.
