# VSI-Super-Wild

From the catalogue. Listed by one researcher from one source page.

- id: `vsi-super-wild`
- kind: other
- robot or device: not stated (class: other)
- how collected: long-form real-world videos with human-verified spatial question answer pairs
- size as stated: 442 view-level videos, 8 scene categories, 6,980 human-verified QA pairs, videos up to 261.08 minutes; 285 hours
- hours: 285 | hours only claimed: none | episodes: not stated
- year: 2026
- organisation: Tsinghua University, NVIDIA, Stanford University
- licence: not stated (class: not stated)
- access: open download (Hugging Face THUSI-Lab/VSI-Super-Wild) (class: open)
- link (dataset): https://vsi-super-wild.github.io/
- note: A third-party list describes the footage as panoramic video taken from YouTube rather than wearer capture; I did not confirm the source.

Confirm the licence at the link before relying on it.