#random sampling

LLMs Fail to Generate Random Numbers From Statistical Distributions

New research reveals large language models cannot faithfully sample from probability distributions, a critical gap for stochastic systems.