If I understand correctly, it's distillation via having a teacher model score each of the student's tokens for a problem based on their own probabilities of generating that token at each step in the sequence. The reward/loss is then applied as RL.
The multi-teacher bit seems to imply they're distilling from multiple models. It's light on the details, but it seems like it could be part of distilling from frontier/closed models. Provided they calculate the logprobs, which OpenAI seems to allow via API but not Anthropic. Maybe they have a way of estimating the logprobs externally?
This method can be used to learn any domain from the teacher. Biology included.
It's not that light on the details. I read the paper and they train several different models in parallel over a few different domains and then they distill from their own models to get the final model.
If I understand correctly, it's distillation via having a teacher model score each of the student's tokens for a problem based on their own probabilities of generating that token at each step in the sequence. The reward/loss is then applied as RL.
The multi-teacher bit seems to imply they're distilling from multiple models. It's light on the details, but it seems like it could be part of distilling from frontier/closed models. Provided they calculate the logprobs, which OpenAI seems to allow via API but not Anthropic. Maybe they have a way of estimating the logprobs externally?
This method can be used to learn any domain from the teacher. Biology included.