x
It seems like misleading responses are mostly not intentional - Benchmarking LLM sophistry — LessWrong