Evaluation Design

AI Reasoning Called into Question After Puzzle Tests and Methods Debate

A recent exchange has shifted the debate from model shortcomings to questions about the tests used to evaluate AI reasoning