ARC-AGI-3 scoring is really weird, in some views it's already saturated in others it's near 0. But I assume, since the entire benchmark IMO is a PR tool for Op

ARC-AGI-3 scoring is really weird, in some views it's already saturated in others it's near 0. But I assume, since the entire benchmark IMO is a PR tool for OpenAI they will publish it eventually.