Large Language Models as Computable Approximations to Solomonoff Induction

Jun Wan, Lingrui Mei

The rapid advancement of large language models (LLMs) calls for a rigorous theoretical framework to explain their empirical success. While significant progress has been made in und…