Blog December 8, 2023 Reducing the Computational Cost of LLMs with Multi-word Tokenization for Sequence Compression
