freePruner selects 50 percent of visual tokens using attention-based importance and keeps accuracy close to the original model, enabling a training-free about 2x acceleration for LMMs.
Token merging: Your ViT but faster
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
freePruner: A Training-free Approach for Large Multimodal Model Acceleration
freePruner selects 50 percent of visual tokens using attention-based importance and keeps accuracy close to the original model, enabling a training-free about 2x acceleration for LMMs.