LLMs Violate Statistical Consistency Principles in Prediction Aggregation17. July 2026AI ModelsLarge language models do not consistently aggregate predictions over subpopulations into valid estimates for overall populations, despite possessing the necessary knowledge. Share on:
