I wonder if the answers from the LLMs could be fed to a single “leader” LLM that could then summarize and determine the most-common answer. I’m guessing you would want one LLM that’s in charge to make the final decision on the answer (or make the call that there is no consensus)?