Fetching the paper…

Flexibly Scaling Large Language Models Contexts Through Extensible Tokenization · Around