inference-paradigm

Recursive Language Models (RLMs) introduce a task-agnostic inference paradigm enabling language models to handle near-infinite contexts by recursively calling themselves over input, with an accompanying open-source inference engine and training environment.