cd/entity/mBART· home entities mBART
grep -l @mbart /news/*.json | wc -l → 1

mBART

mentions 1 type Organization feed RSS

// recent coverage 1 mentions

04:56
2026-07-16
thegustafson.com
natural-language-processing

WordPiece, Unigram, and SentencePiece

Three tokenization algorithms—WordPiece, Unigram, and SentencePiece—each split text differently for language models. WordPiece merges pairs to maximize likelihood, Unigram prunes a large initial vocab…

// co-occurs with top 7 entities