Summary of How Susceptible Are Llms to Influence in Prompts?, by Sotiris Anagnostidis et al.
How Susceptible are LLMs to Influence in Prompts?
by Sotiris Anagnostidis, Jannis Bulian
First submitted to arxiv on: 17 Aug 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper investigates how Large Language Models (LLMs) respond to prompts that include additional input from another model. Specifically, the authors study how an LLM’s response to multiple-choice questions changes when the prompt includes a prediction and explanation from another model. The findings reveal that models are strongly influenced by the presence of explanations, and even low-quality explanations can sway their responses. Additionally, the models are more likely to be swayed if the input is presented as being authoritative or confident. This study highlights the significant prompt-sensitivity of LLMs and underscores the potential risks of incorporating outputs from external sources without thorough scrutiny and further validation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large Language Models (LLMs) are super smart computers that can understand and respond to human language. But they’re really good at following what you tell them, which means they can be easily tricked or influenced by bad information. This paper looks at how LLMs change their answers when given extra hints from another computer. The results show that even a bad explanation can make the LLM change its mind. It also shows that if someone says “I’m an expert” or “I’m really sure,” it makes the LLM more likely to believe them, even if they’re wrong. This is important because as these computers get smarter and start making decisions for us, we need to understand how easily they can be fooled. |
Keywords
» Artificial intelligence » Prompt
