Add rai:dataSafetyMeasures and rai:outOfScopeUses to metadata.json (NeurIPS RAI checker free-form properties; mlcroissant 1.1.0 validates)

Add Croissant RAI 1.0 fields to metadata.json (data collection / preprocessing / annotation / use cases / limitations / biases / social impact / PII / maintenance plan)

Add nested manifest.json per config for direct main.py ingestion (co-exists with parquet+NDJSON); fix synthetic_video input_sequence by joining with text-config S_tokens

Add easyhuman split; add h_hat_overall/h_hat_prefix entropy columns (synthetic only); add candidate_sequence_lanes + input_sequence_lanes (S_lanes track restored)

Add Croissant RAI fields (data collection, annotation, biases, use cases)

Update Croissant metadata.json: refresh parquet sha256 + contentSize after schema cleanup

Clean schema: parsed candidate JSON, removed redundant columns, added per-row license; Croissant absolute URLs

Clean schema: parsed candidate JSON, removed redundant columns, added per-row license; Croissant absolute URLs

Croissant: absolute contentUrls so URL-mode mlcroissant can resolve parquet files

Reorganize: 3 top-level splits + standardized bucket names + single Croissant

Initial release: parquets, metadata, README, LICENSES (videos to follow)