AI benchmarks are a bad joke – and LLM makers are the ones laughing

Study finds many tests don't measure the right things AI companies regularly tout their models' performance on benchmark tests as a sign of technological and intellectual superiority. But those results,…