@block blast While it’s true that many LLMs may contain SAT Math questions in their training datasets, the focus of their training is on understanding patterns and generating coherent responses rather than strictly adhering to any particular educational standard. Excluding certain types of questions for the sake of compliance with Metaculus criteria may not be a priority for developers, especially if those questions contribute to the model's overall performance and versatility in math-related tasks.