This really is an open issue in LLM study without a definite Remedy, but every one of the LLM APIs have an adjustable temperature parameter that controls the randomness in the ou