LLama3 8B vs this model

#1
by danielus - opened

Hello, I'd like to know if this model is better or the DeepMount00/Llama-3-8b-Ita model. I'm not talking about mere mathematical benchmark scores, but rather the feeling while speaking, accuracy in following instructions, and correctness in Italian expressiveness.

I tried the DeepMount00/Mistral-RAG model and, I don't know why, but from time to time it provided Russian outputs.

Thank you so much for what you do! ❤️

Experiment org

Hi @danielus !
Thank you very much for your feedback on DeepMount00/Mistral-RAG. We will update the model promptly to address the unexpected behaviors you've highlighted. Regarding the accuracy in following instructions, the DeepMount00/Llama-3-8b-Ita currently performs better. However, we are actively researching and developing new fine-tuning techniques to enhance the adaptation of open source models to Italian language.

I can't wait! It's really exciting to see that Italy is finally putting its models out there, too. Currently, we see models of all kinds trained on an enormous variety of data, but the only Italian models are very large ones, such as Command R+ and proprietary models.

danielus changed discussion status to closed
danielus changed discussion status to open
DeepMount00 changed discussion status to closed

Sign up or log in to comment