Search papers, labs, and topics across Lattice.
1
0
3
Expert rerouting and attention-aware data packing can boost MoE model throughput by nearly 15%, reshaping efficiency in reinforcement learning.
