Comment on OpenAI o1 Results on ARC-AGI-PubparentComments−krackers1yThere were rumors that 3.5 Sonnet heavily used synthetic data for training, in the same way that OpenAI plans to use o1 to train Orion. Maybe this confirm it?
Comments
There were rumors that 3.5 Sonnet heavily used synthetic data for training, in the same way that OpenAI plans to use o1 to train Orion. Maybe this confirm it?