danny
danjacobson.net
danny
@danjacobson.net
Reading through the ARC-AGI write-up of o3: arcprize.org/blog/oai-o3-...

Does anyone know what the “samples” refer to, specifically? Is it like a N-shot thing, where the model gets 6 or 1024 different cracks at each problem? Or is it more like setting `max_depth` on the CoT?
OpenAI o3 Breakthrough High Score on ARC-AGI-Pub
OpenAI o3 scores 75.7% on ARC-AGI public leaderboard.
arcprize.org
December 28, 2024 at 6:06 PM
Benoit in detroit
November 21, 2024 at 4:07 AM