Search papers, labs, and topics across Lattice.
Zhejiang Sci-Tech University
3
0
5
ClusterMixer reveals that a transparent, hyperbolic clustering approach can significantly outperform traditional opaque token mixers in vision tasks.
Non-uniform bit allocation can boost recall by up to 18% in low-bit quantization, reshaping how we approach embedding compression.
SOPD not only outperforms traditional distillation methods but also redefines how we think about trajectory corrections in model training.
