MultiTok: Variable-Length Tokenization for Efficient LLMs Adapted from LZW Compression

TL;DR

MultiTok introduces a variable-length tokenization method inspired by LZW compression, enabling more efficient training of large language models with comparable accuracy and significantly reduced resource requirements.

Contribution

The paper presents MultiTok, a novel tokenization approach that improves training efficiency and reduces data needs for large language models.

Findings

MultiTok achieves 2.5x faster training times.

It requires over 30% less training data.

Performance is comparable to BERT and GPT tokenizers.

Abstract

Large language models have drastically changed the prospects of AI by introducing technologies for more complex natural language processing. However, current methodologies to train such LLMs require extensive resources including but not limited to large amounts of data, expensive machinery, and lengthy training. To solve this problem, this paper proposes a new tokenization method inspired by universal Lempel-Ziv-Welch data compression that compresses repetitive phrases into multi-word tokens. With MultiTok as a new tokenizing tool, we show that language models are able to be trained notably more efficiently while offering a similar accuracy on more succinct and compressed training data. In fact, our results demonstrate that MultiTok achieves a comparable performance to the BERT and GPT standards as both a stand-alone tokenizer and an add-on to existing tokenizers while also providing close to 2.5x faster training with more than 30% less training data.