Journey9ni/vstibench

VSTI-Bench VSTI-Bench is a benchmark for visual-spatial temporal intelligence, evaluating video models on camera- and object-centric spatial reasoning over ScanNet videos. It accompanies the VLM-3R project. Videos: ScanNet.tar.gz (val-split ScanNet videos) Questions: test.json (5,736 QA items across 9 question types) Erratum (2026-07-13): corrected camera-position ground truth A bug in the ground-truth generation pipeline was reported by Jacob Yeung (CMU), whom… See the full description on the dataset page: https://huggingface.co/datasets/Journey9ni/vstibench.

<iframe src="https://silytics.si/embed/item/a454a1a5-1804-4409-837a-f2d41c2bd330" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.