METok: Multi-Stage Event-based Token Compression for Efficient Long Video Understanding
This work addresses efficiency issues in long video understanding for VLLM users, representing an incremental improvement through a novel compression method.
The paper tackles the challenge of high computational demands and redundancy in processing long videos for Video Large Language Models (VLLMs) by proposing METok, a training-free token compression framework that achieves an 80.6% FLOPs reduction and 93.5% KV Cache memory savings while maintaining comparable or superior accuracy.
Recent advances in Video Large Language Models (VLLMs) have significantly enhanced their ability to understand video content. Nonetheless, processing long videos remains challenging due to high computational demands and the redundancy present in the visual data. In this work, we propose METok, a training-free, Multi-stage Event-based Token compression framework designed to accelerate VLLMs' inference while preserving accuracy. METok progressively eliminates redundant visual tokens across three critical stages: (1) event-aware compression during vision encoding, (2) hierarchical token pruning in the prefilling stage based on semantic alignment and event importance, and (3) a decoding-stage KV Cache optimization that further reduces memory consumption. Our experiments on diverse video benchmarks demonstrate that METok achieves an optimal trade-off between efficiency and accuracy by dynamically selecting informative visual tokens. For instance, equipping LongVA-7B with METok realizes an 80.6% FLOPs reduction and 93.5% KV Cache memory savings, all while maintaining comparable or even superior accuracy.