Based-30b
https://huggingface.co/ehartford/based-30b
So, as I was working on Wizard-Vicuna-30b-Uncensored and WizardLM-Uncensored-Falcon-7b, I came to the realization that these models, despite being trained with no refusals, were still refusing. How could thi...
erichartford.com6 min read
pepe
This is very insightful to get a better understanding of the underlying model. It would be interesting to try this with ChatGLM, it's a Chinese (yet bilingual) model. It would be interesting to see if there are any differences based on what country the LLM is made in. Especially since a lot of websites are banned in China. Although, I'm not sure if that affected the data collection process.