Eh, I've found llama 7b finetunes to be underwhelming too. They both seem to fall apart at detail retrieval and such, even with Mistral's longer base context.

Eh, I've found llama 7b finetunes to be underwhelming too. They both seem to fall apart at detail retrieval and such, even with Mistral's longer base context. For what its worth, Mistral finetunes excel in this ranking, and Mistral AI almost certainly did not contaminate it with that test: https://rentry.co/ayumi_erp_rating https://rentry.co/ayumi_erp_rating