RLM enables LLMs to process arbitrarily long contexts by treating them as an external environment that can be programmatically explored via a JavaScript REPL. Instead of feeding entire long documents ...