Dynamically selecting the most diverse ensemble set of multiple LLMs and fusing the outputs with Multi Agent Reinforcement Learning. Overal system can switch between models according to the tasks and boost the overall accuracy and robustness - View it on GitHub