[alibero]

Check out figures 1 & 2 in the Llama-2 paper :) They benchmark against ChatGPT for helpfulness and harmfulness

https://ai.meta.com/research/publications/llama-2-open-found...