An intelligence quotient (IQ) test measures performance on a finite set of capabilities, often specific logic puzzles completed under time constraints. It can correlate with outcomes such as school performance, but it does not provide an objective definition of intelligence or reduce intelligence to a single underlying cause. section 1.4
Different IQ tests exist because intelligence is difficult to define. Their usefulness comes from repeated refinement and from identifying answers that correlate with human performance—not from directly measuring the cause of that performance. section 7.1.1 section 1.4
Using IQ tests to judge an LLM
An LLM may perform impressively on an IQ test, just as it may perform well on the MCAT or bar exam. That result does not establish human-like intelligence: training data may contain similar questions, and IQ tests were designed to predict aspects of human performance rather than evaluate AI algorithms. section 7.1.1 section 1.4
