Comment Comment by michaelhartm They used the 6b GPT4-J, not 20B. That's what's interesting, it's a smallish large language model :). parent michaelhartm 3 years ago
