Or maybe the temperature is set too high! Probably don't start with a single LLM response as the basis for your understanding of scientific consensus.