ClockBench AI Benchmark
clockbench.ai
ClockBench evaluates whether models can read analog clocks - a task that is trivial for humans, but current frontier models struggle with.
  • hendrik@palaver.p3x.deEnglish
    3·
    3 个月前

    The human baseline is nice as well. 11% of people can’t read analog clocks either. And reportedly it goes way down with younger people. But yeah, AI seems to be abysmal.