Time Blindness: Why Video-Language Models Can't See What Humans Can?
This addresses a fundamental limitation in video understanding for AI applications, exposing a gap between human and machine temporal reasoning that could hinder progress in fields like surveillance or medical imaging.
The paper tackles the problem of video-language models' inability to capture purely temporal patterns when spatial information is obscured, revealing that while humans achieve over 98% accuracy on the SpookyBench benchmark, state-of-the-art models score 0% accuracy. It shows that models degrade more rapidly than humans in low spatial signal-to-noise ratio tasks, highlighting a critical reliance on spatial features.
Recent advances in vision-language models (VLMs) have made impressive strides in understanding spatio-temporal relationships in videos. However, when spatial information is obscured, these models struggle to capture purely temporal patterns. We introduce $\textbf{SpookyBench}$, a benchmark where information is encoded solely in temporal sequences of noise-like frames, mirroring natural phenomena from biological signaling to covert communication. Interestingly, while humans can recognize shapes, text, and patterns in these sequences with over 98% accuracy, state-of-the-art VLMs achieve 0% accuracy. This performance gap highlights a critical limitation: an over-reliance on frame-level spatial features and an inability to extract meaning from temporal cues. Furthermore, when trained in data sets with low spatial signal-to-noise ratios (SNR), temporal understanding of models degrades more rapidly than human perception, especially in tasks requiring fine-grained temporal reasoning. Overcoming this limitation will require novel architectures or training paradigms that decouple spatial dependencies from temporal processing. Our systematic analysis shows that this issue persists across model scales and architectures. We release SpookyBench to catalyze research in temporal pattern recognition and bridge the gap between human and machine video understanding. Dataset and code has been made available on our project website: https://timeblindness.github.io/.