EvoPrune: Early-Stage Visual Token Pruning for Efficient MLLMs
EvoPrune addresses the problem of inefficient MLLM inference for applications requiring low latency, such as real-time video processing.
The authors tackled the problem of inefficient inference in Multimodal Large Language Models (MLLMs) due to exponential growth of visual tokens, and achieved a 2x inference speedup with less than 1% performance degradation using their proposed EvoPrune method. This was demonstrated on the VideoMME dataset.
Multimodal Large Language Models (MLLMs) have shown strong performance in vision-language tasks, but their inference efficiency is severely limited by the exponential growth of visual tokens in complex scenarios such as high-resolution images and videos. Existing visual token pruning methods mainly operate after visual encoding, overlooking the substantial computational cost incurred during the encoding stage. To address this issue, we propose EvoPrune, an early-stage visual token pruning method for MLLMs that performs pruning directly during visual encoding. Specifically, EvoPrune employs a layer-wise pruning strategy guided by token similarity, diversity, and attention-based importance to retain the most informative visual tokens at selected encoding layers. Extensive experiments on image and video benchmarks validate the effectiveness of EvoPrune. In particular, on the VideoMME dataset, EvoPrune achieves 2$\times$ inference speedup with less than 1% performance degradation, demonstrating its potential for latency-sensitive MLLM deployment.