When we hear the word “uncensored” in AI, we think of models that can be harmful and possibly biassed. While that can be absolutely true, these uncensored models are increasingly outperforming their aligned counterparts, even if possibly harmful.
According to the Open LLM Leaderboard on Hugging Face, uncensored models like Wizard-Vicuna-13B-Uncensored-HF, whose developer recently got into trouble for releasing the model to the public, have been competing with LLaMa and…
Read the full article here