Sample (20 rows per task): metadata.parquet, audio/, README, LICENSE (CC BY-NC-SA 4.0)

Remove metadata.parquet before sample refresh

Upload sample: metadata.parquet + audio/ (20 rows per subset)

Add BEANS-Next dataset card (tier/subset layout)

Remove dataset data files (keep Croissant + datasheet)

Upload BEANS-Next Parquet tables (exclude unseen genus/family 4way)

Remove previous Parquet shards

Upload BEANS-Next Parquet tables (union schema, 20/sample)

Remove previous Parquet shards

Upload BEANS-Next Parquet tables (20/sample)

Remove previous Parquet shards

beans-next: two-table Parquet bundle (20/subset, tiers 1–4, deduped audio)

Remove previous Parquet shards

Add BEANS-Next dataset card (audio + metadata tables)

beans-next: two-table Parquet smoke (5/subset tier1 + tier4 minus unseen)

Remove previous Parquet shards

Add BEANS-Next dataset card (tier/subset layout)

Add BEANS-Next dataset card (tier/subset layout)

Add BEANS-Next dataset card (tier/subset layout)

Add small tier_1/zebra-description shard (32 rows)

Add BEANS-Next dataset card (tier/subset layout)

Add tier_1/crow-description test.parquet

Add BEANS-Next dataset card (tier/subset layout)

initial commit