Why is it that with ChatGPT, Gemini, Claude, Copilot and other LLMs we have to always start new chats for them to work well? What is the scientific explanation? What are the hypotheses? What is the evidence for each?
Why LLMs Require New Chats for Optimal Performance
By
–