Implementation Mm-cot

[Billyroot]

Great work from yourself and your team. Quick question,we are thinking about using that method with a largeur Falcon model. do you think we Can have therefore a greater gap of performance with gpt 3.5?the idea being if a 1b model Can Do that, what Can be with a 40b model.

[cooelf]

Not sure about that. However we did see that when using a T5-style encoder-decoder model, a larger model achieves better performance. Due to the resource limit, we did not scale to models larger than 1B.