Comment on Perplexity open sources R1 1776, a version of the DeepSeek R1 model that CEO Aravind Srinivas says has been “post-trained to remove the China censorship”.
LLMs are gigantic bias reproduction machines, so it will never be perfect.