论文里程碑 #22:Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates2026-10-28·5207 字·11 分钟LLM LLM 论文里程碑 Unigram 子词正则化 SentencePiece 神经机器翻译
11 Unigram(SentencePiece):并非所有切分都是确定的2026-06-09·4006 字·8 分钟LLM LLM Tokenizer Unigram SentencePiece Subword Regularization EM Viterbi 分词