Ai2 Helps Build 47B-Token Thai Corpus Mangosteen

Thai researchers used Ai2's Dolma toolkit to build Mangosteen, a 47-billion-token Thai pretraining corpus. Models trained on it matched or beat those trained on larger web corpora after filtering low-quality and duplicate text.

2026-08-27 ~ 2026-08-27 · 2 related posts

1 near-duplicate retellings: allen_ai