This work attempts to use various prompts with different seed values passed to the LLM. Includes response accuracy and generating harmful responses. Paper: https://www.ijsat.org/papers/2023/3/1126.pdf