ResearchBenchmarksSentence Retrieval on 20 languages Same scriptFollow49.1AccuracyBPE44.21245.48146.7548.019May 26, 2023Evaluation ResultsMethodMethodLinksAccuracyBPETokenizer=BPETokenizer=BPE2023.0549.1TokMixTokenizer=TokMixTokenizer=TokMix2023.0546.9UnigramTokenizer=UnigramTokenizer=Unigram2023.0544.4