FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams?

Key points

• Under bounded context budgets, models must balance temporal history, spatial resolution, and temporal granularity; sparse sampling at 1--2 FPS misses fast events.1
• We introduce FastBench to evaluate high-dynamic perception in real-world video streams.1
• We also present ProactiveFrame, a training-free baseline that adjusts incoming frame rates through text tokens.1
• FastBench provides a testbed for high-dynamic streaming video understanding.1

Sources (1)

• [1]FastBench: Can Streaming VLMs Perceive High-Dynamic Real-World Streams?
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 05:55 PM

  Streaming Video Large Language Models (VLMs) enable continuous video understanding, yet existing benchmarks focus on low-dynamic scenarios.

  Under bounded context budgets, models must balance temporal history, spatial resolution, and temporal granularity; sparse sampling at 1--2 FPS misses fast events.

Extractive summary: sentences quoted from the sources.