FLEXITOKENS: Flexible Tokenization for Evolving Language Models
This addresses the challenge of adapting language models to evolving data, such as new languages or domains, with incremental improvements over existing tokenizer-free methods.
The paper tackled the problem of rigid subword tokenizers in language models, which cause inefficient tokenization when adapting to new data distributions, and introduced FLEXITOKENS, a method with a simplified training objective that reduces token over-fragmentation and achieves up to 10% improvements on downstream task performance.
Language models (LMs) are challenging to adapt to new data distributions by simple finetuning. This is due to the rigidity of their subword tokenizers, which typically remain unchanged during adaptation. This inflexibility often leads to inefficient tokenization, causing overfragmentation of out-of-distribution domains, unseen languages, or scripts. In this work, we develop byte-level LMs with learnable tokenizers to make tokenization adaptive. Our models include a submodule that learns to predict boundaries between the input byte sequence, encoding it into variable-length segments. Existing tokenizer-free methods train this boundary predictor using an auxiliary loss that enforces a fixed compression rate across the training corpus, introducing a new kind of rigidity. We propose FLEXITOKENS, a simplified training objective that enables significantly greater flexibility during adaptation. Evaluating across multiple multilingual benchmarks, morphologically diverse tasks, and domains, we demonstrate that FLEXITOKENS consistently reduces token over-fragmentation and achieves up to 10% improvements on downstream task performance compared to subword and other gradient-based tokenizers. Code and data for our experiments will be released at https://github.com/owos/flexitokens