MI-Pruner prunes visual tokens in MLLMs using crossmodal mutual information computed prior to feature interaction and outperforms attention-based pruning methods with minimal latency.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
MI-Pruner: Crossmodal Mutual Information-guided Token Pruner for Efficient MLLMs
MI-Pruner prunes visual tokens in MLLMs using crossmodal mutual information computed prior to feature interaction and outperforms attention-based pruning methods with minimal latency.