Search papers, labs, and topics across Lattice.
University of Central Florida
2
0
4
DnA achieves a notable 0.8% improvement on ImageNet-1K by effectively filtering out irrelevant features in attention mechanisms.
Pruning VLMs doesn't have to destroy chain-of-thought reasoning – MuCRASP shows how to surgically remove parameters while maintaining reasoning consistency and even outperforming strong baselines.
