“Tokenizer Choice For LLM Training: Negligible or Crucial?”, Mehdi Ali, Michael Fromm, Klaudia Thellmann, Richard Rutmann, Max Lübbering, Johannes Leveling, Katrin Klug, Jan Ebert, Niclas Doll, Jasper Schulze Buschhoff, Charvi Jain, Alexander Arno Weber, Lena Jurkschat, Hammam Abdelwahab, Chelsea John, Pedro Ortiz Suarez, Malte Ostendorff, Samuel Weinbach, Rafet Sifa, Stefan Kesselheim, Nicolas Flores-Herr2023-10-12 (LM tokenization, GPT-2)⁠:

The recent success of LLMs (Large Language Models) has been predominantly driven by curating the training dataset composition, scaling of model architectures and dataset sizes and advancements in pretraining objectives, leaving tokenizer influence as a blind spot. Shedding light on this underexplored area, we conduct a comprehensive study on the influence of tokenizer choice on LLM downstream performance by training 24 mono- and multilingual LLMs at a 2.6b parameter scale, ablating different tokenizer algorithms and parameterizations.

Our studies highlight that the tokenizer choice can impact the model’s downstream performance, training and inference costs. In particular, we find that the common tokenizer evaluation metrics fertility and parity are not always predictive of model downstream performance, rendering these metrics a questionable proxy for the model’s downstream performance.

Furthermore, we show that multilingual tokenizers trained on the 5 most frequent European languages require vocabulary size increases of factor 3 in comparison to English. While English-only tokenizers have been applied to the training of multi-lingual LLMs, we find that this approach results in a severe downstream performance degradation and additional training costs of up to 68%, due to an inefficient tokenization vocabulary.