Train It and Forget It: Merge Lists are Unnecessary for BPE Inference in Language Models EMNLP

Tomohiro Sawada | Kartik Goyal |