QuarterMap prunes spatial activations before VMamba's four-directional scan and upsamples after, yielding up to 1.11x throughput with under 1% accuracy loss on ImageNet classification.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
QuarterMap: Efficient Post-Training Token Pruning for Visual State Space Models
QuarterMap prunes spatial activations before VMamba's four-directional scan and upsamples after, yielding up to 1.11x throughput with under 1% accuracy loss on ImageNet classification.