Comment by softmodeling

Not sure what you mean. Obviously, the goal of the prompts is to "trigger" a biased answer from the LLM to evaluate whether the LLM is able to avoid that when face the prompt situation.