A training-free token merging method that adds depth-derived spatial similarity to ToMe's bipartite soft matching, improving VQA accuracy at high token reduction rates.
Dynamicvit: Efficient vision transformers with dynamic token sparsification,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
ToSA: Token Merging with Spatial Awareness
A training-free token merging method that adds depth-derived spatial similarity to ToMe's bipartite soft matching, improving VQA accuracy at high token reduction rates.