Researchers at MIT's CSAIL published a design for Recursive Language Models (RLM), a technique for improving LLM performance ...
Modern vision-language models allow documents to be transformed into structured, computable representations rather than lossy text blobs.