thorn turret
A MoE AI (Mixture of Experts) consists on a AI that is trained to select one LLM among others (or even another MoE AI) to predict the next token.
Seems possible for me that Vedal added more LLM's on Neuro-Sama from time to time and united everything in one fundamental MoE AI, so Neuro-Sama have actual continuity in her old internal model traits even though she is smarter because of the new LLM's.
inland geyser