SozKZ Corpora: Kazakh Training Datasets Collection Training corpora for Kazakh LLMs — raw, cleaned, deduplicated, tokenized, synthetic, and parallel datasets • 14 items • Updated Jun 11 • 1