card: C1 composition + wongnai attribution

round C1: v2-exact + wongnai 6k (2k/2k/2k)

b3: dataset card refresh (31,091 rows)

b3: v2-exact base 31,091 rows (revert MASSIVE-th + oversample)

card: trainset.jsonl = 32,638 rows (B2 round, 2026-09-25)

B2: trainset.jsonl -> 32,638 rows (v2 + MASSIVE-th 1,547, no oversample)

Upload trainset.jsonl with huggingface_hub

Upload trainset.jsonl with huggingface_hub

Add trainset.jsonl (29,667 rows, seed 42)

Add dataset card

initial commit