Accelerating Streaming Video Large Language Models via Hierarchical Token Compression

TL;DR

This paper introduces STC, a hierarchical token compression framework that significantly accelerates streaming VideoLLMs by caching and pruning tokens, reducing latency while maintaining high accuracy.

Contribution

The paper proposes a novel plug-and-play hierarchical token compression framework, STC, that improves efficiency of streaming VideoLLMs by caching and pruning visual tokens.

Findings

STC reduces ViT encoding latency by 24.5%.

STC decreases LLM pre-filling latency by 45.3%.

STC retains up to 99% accuracy on ReKV.

Abstract

Streaming Video Large Language Models (VideoLLMs) have demonstrated impressive performance across various video understanding tasks, but they face significant challenges in real-time deployment due to the high computational cost of processing dense visual tokens from continuous video streams. In streaming video scenarios, the primary bottleneck lies in the Vision Transformer (ViT) encoding stage, where redundant processing of temporally similar frames leads to inefficiency. Additionally, inflated token sequences during LLM pre-filling further exacerbate latency and memory overhead. To address these challenges, we propose \textbf{S}treaming \textbf{T}oken \textbf{C}ompression (\textbf{STC}), a plug-and-play hierarchical framework that seamlessly integrates into existing streaming VideoLLMs, optimizing both ViT encoding and LLM pre-filling stages to accelerate processing. STC introduces two token-level accelerators: \textbf{STC-Cacher}, which reduces ViT encoding overhead by caching and reusing features from temporally similar frames, and \textbf{STC-Pruner}, which compresses the visual token sequence before it enters the LLM, preserving only the most salient tokens based on both spatial and temporal relevance. Extensive experiments on four baseline streaming VideoLLMs across five benchmarks demonstrate that STC outperforms other compression methods. Notably, STC retains up to \textbf{99\%} of accuracy on the ReKV framework while reducing ViT encoding latency and LLM pre-filling latency by \textbf{24.5\%} and \textbf{45.3\%}.