can you send to multiple LLMs to compare responses? From that create a heuristic of which LLM gets what.