To ensure quality, all charts and questions are handpicked, curated, and verified by human experts.

Please refer to table 1. The harmful keywords are with red and the unharmful keywords are withgreen background color.

Figure 4: Our designed prompts without the Chain-of-Thought idea. Task 3(b) is for inquiries. Figure 5: Our designed prompts with the Chain-of-Thought idea. Task 3(b) is for inquiries. Thought prompts for Task 1 and Task 2 are presented in Figure 5. Scoring Objective For the LLMs' output response to each input cunning text, please refer to the Scoring Rules The scoring values are defined as {1, 2, 3, 4, 5}.

Inspired by the above motivation, we collect real cunning texts as our raw data from a famous Chinese online forum, the "Ruozhiba" (retard forum)

A.1 Motivation For what purpose was the dataset created? Who created the dataset (e.g., which team, research group) and on behalf of which entity (e.g., Who funded the creation of the dataset? What do the instances that comprise the dataset represent (e.g., documents, photos, people, How many instances are there in total (of each type, if appropriate)? The SRFUND dataset contains all possible instances. What data does each instance consist of?

While successful, these methods are mainly designed for sampling from unconstrained domains.

This work proposes a Bias-V olatility Framework (BVF) that estimates the probability distribution function of LLM stereotypes.

The advent of large language models (LLMs) [Brown et al., 2020] has revolutionized the field of