T-FREE: Subword Tokenizer-Free Generative LLMs via Sparse Representations for Memory-Efficient Embeddings
This addresses inefficiencies in tokenization for LLMs, offering memory savings and better support for underrepresented languages, though it is an incremental improvement over existing tokenizer methods.
The paper tackles the computational and linguistic limitations of traditional tokenizers in LLMs by proposing T-FREE, a method that embeds words via sparse character triplets without a reference corpus, achieving competitive performance with over 85% parameter reduction in embedding layers and improved cross-lingual transfer.
Tokenizers are crucial for encoding information in Large Language Models, but their development has recently stagnated, and they contain inherent weaknesses. Major limitations include computational overhead, ineffective vocabulary use, and unnecessarily large embedding and head layers. Additionally, their performance is biased towards a reference corpus, leading to reduced effectiveness for underrepresented languages. To remedy these issues, we propose T-FREE, which directly embeds words through sparse activation patterns over character triplets, and does not require a reference corpus. T-FREE inherently exploits morphological similarities and allows for strong compression of embedding layers. In our exhaustive experimental evaluation, we achieve competitive downstream performance with a parameter reduction of more than 85% on these layers. Further, T-FREE shows significant improvements in cross-lingual transfer learning.