AI benchmarks hampered by bad science

: Study finds many tests don’t measure the right things

Read in full here: