ReDiPrune: Relevance-Diversity Pre-Projection Token Pruning for Efficient Multimodal LLMs

Abstract: Recent multimodal LLMs are computationally expensive because Transformers must process a large number of visual tokens. We present \textbf{ReDiPrune}, a training-free token pruning method applied before the vision-language projector, where visual features remain rich and discriminative. Unlike post-projection pruning methods that operate on compressed representations, ReDiPrune selects informative tokens directly from vision encoder outputs, preserving fine-grained spatial and semantic cues. Each token is scored by a lightweight rule that jointly consider text-conditioned relevance and max-min diversity, ensuring the selected tokens are both query-relevant and non-redundant. ReDiPrune is fully plug-and-play, requiring no retraining or architectural modifications, and can be seamlessly inserted between the encoder and projector. Across four video and five image benchmarks, it consistently improves the accuracy-efficiency trade-off. For example, on EgoSchema with LLaVA-NeXT-Video-7B, retaining only 15\% of visual tokens yields a +2.0\% absolute accuracy gain while reducing computation by more than $6\times$ in TFLOPs. Code is available at https://github.com/UA-CVML/ReDiPrune.