Understanding Recursive Language Models: Managing Long Contexts
A deep dive into RLMs - a new inference paradigm that lets LLMs process millions of tokens by treating context as a part of an external environment which a LLM can interact with. Includes a minimal implementation from scratch.
Read more