Language models can largely benefit from efficient tokenization. However, they still mostly utilize the classical BPE algorithm, a simple and reliable method. This has been shown to cause such issues as under-trained tokens and sub-optimal compression that may affect the downstream performance. We introduce Picky BPE, a modified BPE algorithm that carries out vocabulary refinement during tokenizer training. Our method improves vocabulary efficiency, eliminates under-trained tokens, and does not compromise text compression. Our experiments show that our method does not reduce the downstream performance, and in several cases improves it.
@article{arxiv.2409.04599,
title = {BPE Gets Picky: Efficient Vocabulary Refinement During Tokenizer Training},
author = {Pavel Chizhov and Catherine Arnett and Elizaveta Korotkova and Ivan P. Yamshchikov},
journal= {arXiv preprint arXiv:2409.04599},
year = {2024}
}