Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
3
0
3
Recursive refinement unlocks state-of-the-art spatial context prediction with just 11.2M trainable parameters, outperforming larger single-step models by 44%.
Achieving 3.1 TOPS/W energy efficiency, Chimera sets a new benchmark for ultra-low-power AI inference at the edge.
Unlock up to 7x speedup and 4.9x energy efficiency in MX-MatMul with VMXDOTP, a RISC-V Vector ISA extension that finally makes microscaling data formats practical for vector processors.