Rude prompts to LLMs consistently lead to better results than polite ones 🤯
The authors found that very polite and polite tones reduced accuracy, while neutral, rude, and very rude tones improved it.
Statistical tests confirmed that the differences were significant, not random, across repeated runs.
The top score reported was 84.8% for very rude prompts and the lowest was 80.8% for very polite.
They compared their results with earlier studies and noted that older models (like GPT-3.5 and Llama-2) behaved differently, but GPT-4-based models like ChatGPT-4o show this clear reversal where harsh tone works better.
----
Paper – arxiv. org/abs/2510.04950
Paper Title: "Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy (short paper)"
Andy Jassy, the current CEO of Amazon, has been with the company for 28 years. In 2003, he played a pivotal role in creating and scaling AWS which is the most valuable part of Amazon. A significant portion of the modern internet is powered by AWS.
Amazon today is worth $2.4 Trillion and is one of the most critical companies to the US and the world. Yet the CEO makes less money than many YouTubers, comedians, podcasters, professional athletes, and actors, that don't manage or guide anything nearly as important or critical to the US or have millions of employees.
We need to stop pretending that the CEO of fortune 10 companies who helped build trillions of dollars worth of values for investors should have their salary compared with the delivery drivers working their way through college. This trope is old, and should be retired.
Jeff Bezos believes that AI is an “industrial bubble”, which means every type of company or experiment — good idea or bad idea — will get funded.
Even if investors don’t make money, AI is very “real” and “will change every industry”.
He cites biotech bubble in 1990s: investors as whole lost money but society gained a number of life-saving drugs (similarly, the telecom bubble resulted in a massive cable buildout we still use).