Making AI benchmarks harder without accidentally adding bias

Original: @cwolferesearch Interesting point, I’ve found the hardest part is defining harder without bias.

Source: x.com ↗