viewer: add configs YAML + metadata.jsonl for 2dbbox & dense_captioning (#22)

Dataset card refresh + promote CHANGELOG.md as primary release history (#20)

README: replace dataset_info with per-task configs so HF viewer loads each schema independently (#21)

VANTAGE-Bench update: prompts + question-only annotations + Vantage2DPointing rename + test split + data/README + CHANGELOG (#19)

Declare test-only split in dataset_info frontmatter

Point License section to LICENSE.md in dataset

Remove paper citation TBD reminder

Rename LICENSE -> LICENSE.md so it resolves from frontmatter

croissant: ensure subset has the latest version with license/citeAs/version/datePublished/RAI (#17)

Rename LICENSE.md -> LICENSE (#16)

croissant: keep only rai:-prefixed RAI fields (deduplicate) (#14)

croissant.json: cumulative update (RAI dual keys, citation, new title/authors) (#12)

croissant: dual rai: prefixed + unprefixed RAI keys for validator compatibility (#9)

croissant.json: unprefix RAI fields, add @context aliases (NeurIPS RAI compatibility) (#8)

croissant.json: split aggregated recordSets to one per fileset (source cardinality fix) (#7)

Update Croissant 1.1 metadata: fix forbidden chars, paths, recordSet IDs (#6)

Add Croissant 1.1 metadata (croissant.json) (#4)

Restructure data/: drop Spatial, rename to dense_captioning / event_verification / pointing / referring / temporal_localization / vqa (#5)

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