Comment by nearbuy

They would score much worse on the private set than the public set. And they haven't done this for any of the other ARC-AGI benchmarks, so why would they do it for this one?