Very much appreciate the work that’s gone into this model. After a little look around the repo and announcement it’s unclear if they’ve published the post-train

Very much appreciate the work that’s gone into this model. After a little look around the repo and announcement it’s unclear if they’ve published the post-training data set they used. If they have published the post-training it’ll go a long way towards completing the value prop of this model. If not, it is hard to know whether they’ve replaced one set of biases and refusals with another. The message “trust us! We are American!” is strong - but showing the data is stronger.