Search papers, labs, and topics across Lattice.
Renmin University of China
3
0
6
EntroRouter achieves near-optimal expert accuracy while slashing computational costs by nearly half, challenging the status quo of model routing efficiency.
iLLaDA's fully bidirectional diffusion training outperforms traditional autoregressive models, achieving remarkable gains across key language benchmarks.
Aligning router rows with principal singular directions significantly boosts the effectiveness of Mixture-of-Experts models, challenging existing design principles.
