Add Rosettia YouTube entry

Add author credits (Estefania Espinosa Fernandez & Irving Ernesto Quezada Ramirez) + contributions

Safety remediation: redistribute ONLY the Manual-derived Chanka data (CC-BY-4.0, attribution to Min. de Cultura/Ardito Vega). Remove JW300/AmericasNLP, SomosNLP, FLORES, Benito-2018, Nouman/RunaSimi, C4-synth and eval outputs (prohibited/viral/unconfirmed licenses) - documented + reconstructable via scripts. Honest provenance card.

Upload eval_mbr/v30_self_dedup.json with huggingface_hub

Upload eval_mbr/v30_mbr_n32.json with huggingface_hub

Upload eval_mbr/nllbv2_self_mbr.json with huggingface_hub

Upload eval_mbr/gspo_self_mbr.json with huggingface_hub

Upload eval_mbr/ens_v30_v32.json with huggingface_hub

Upload eval_mbr/ens_v30_nllbv2.json with huggingface_hub

Upload eval_mbr/ens_v30_nllb.json with huggingface_hub

Upload eval_mbr/ens_v30_gspo.json with huggingface_hub

Upload eval_mbr/ens_strong3.json with huggingface_hub

Upload eval_mbr/ens_4way.json with huggingface_hub

Upload eval_mbr/ens_3way_madlad.json with huggingface_hub

Upload eval_mbr/abl_nllbr2_val.json with huggingface_hub

Upload eval_mbr/gspo_cands_n64.jsonl with huggingface_hub

Upload eval_mbr/nllb_v2_cands_n64.jsonl with huggingface_hub

Upload eval_mbr/v30_cands_n64.jsonl with huggingface_hub

Upload synth/synth_v30_200k.jsonl with huggingface_hub

Upload rl/nllb_aggregate_corpus.parquet with huggingface_hub

Upload rl/nllb_v2_corpus.parquet with huggingface_hub

Upload rl/rl_val_split.parquet with huggingface_hub

Upload rl/rl_train_split.parquet with huggingface_hub

Upload rl/rl_heldout_quy.parquet with huggingface_hub

Upload clean_chanka/v34_raw_corpus.parquet with huggingface_hub

Upload clean_chanka/v34_training_corpus_clean3.parquet with huggingface_hub

Upload clean_chanka/v34_training_corpus.parquet with huggingface_hub

v32: v30 + Nouman sentence-level quy (no dict_misc/RunaSimi after v31 lesson)

Add v31 expanded Chanka corpus (15,670 pairs: v30 + Nouman 2023 quy + RunaSimi.de Ayakuchu)

Add v30 expanded Chanka training corpus (+874 pairs from glossary/Benito merge)

Remove synthetic AmericasNLP manifest row

Document real-only AmericasNLP tier

Remove synthetic AmericasNLP public artifact

Add AmericasNLP Quechua-Spanish tier

Initial RosettIA Chanka dataset release

initial commit