Tag
Miles Brundage highlights excellent work from the AI Security Institute investigating the impact of test-time compute budgets for frontier AI model evaluations, with praise from Noam Brown.