SOURCE-LINKED INTELLIGENCE

ShallowStream: Index Shallow then Answer Deep for Streaming Video Understanding

arXiv · AI, language, vision and robotics · article · Sep 2, 2026 · UTC

Streaming video understanding is a critical capability for real-world applications, including embodied intelligence, autonomous driving, industrial monitoring, surveillance and early warning, and wearable assistants. However, processing continuous video streams with multimodal large language models (MLLMs) is computationally expensive. Existing efforts have explored reducing streaming overhead through visual token pruning, token merging, quantization, on-demand frame retrieval, and context offloading. However, most existing methods overlook the dimension of model depth. Repeatedly executing fu

Read original source ↗ Open in workspace

recordType

paper

region

Global