Research
5h agoApple Finds LLMs Fail Bayesian Belief Updates
Researchers at Apple discovered that LLMs fail to update probabilistic beliefs consistently when presented with new evidence. The team measured this using a novel information processing gap to quantify deviations from Bayes' rule. These internal inconsistencies hinder rational decision-making in high-stakes fields.
The Signal
Practitioners should avoid relying on LLMs for precise probabilistic reasoning in medicine or law.