Restructure data/: drop Spatial, rename to dense_captioning / event_verification / pointing / referring / temporal_localization / vqa

Replace data/ with new 4 GB proportional subset (#3)

Remove annotations/ — subset hosts media only (#2)

Move task overview figure from Dataset Description to References

Mirror VANTAGE-Bench dataset card and add markdown LICENSE

Add VANTAGE-Bench subset (media + annotations + MANIFEST) (#1)

initial commit