Yes, but I meant it slightly differently than the distills.
The idea is to create the next gen SOTA non reasoning model with synthetic reasoning training data.
The idea is to create the next gen SOTA non reasoning model with synthetic reasoning training data.