# Byte-level processing

MrT5 is an efficient improved version of ByT5. It shortens the input sequence length through dynamic token merging technology to improve processing efficiency.

ByT5 is Google's tokenizer-free version of T5, directly processing UTF-8 byte sequences with native multilingual text handling, especially excelling at noisy data.

ByT5 is a tokenizer-free version of Google's T5 that directly processes raw UTF-8 bytes without relying on a tokenizer, making it particularly suitable for handling noisy text and multilingual scenarios.

ByT5 is a tokenizer-free version of Google's T5 that directly processes raw UTF-8 bytes, supporting multilingual text processing with excellent performance on noisy data.

ByT5 is a tokenizer-free version of Google's T5 that directly processes UTF-8 byte sequences, supporting multilingual text processing with robustness to noisy data.