[ New! ] “Pre-trained Large Language Models Learn to Predict Hidden Markov Models In-context” joint work with Y. Dai, Z. Gao, S. Dean, and J. Sun is now accepted to NeurIPS 2025! (paper, github)