Search papers, labs, and topics across Lattice.
Peking University
1
0
3
Hypic slashes time-to-first-token by 2.45x and doubles throughput for hybrid-attention LLMs, all while preserving near-full accuracy.
