HackerHACKER PAPER · 2025 · arXiv preprint
Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy
这篇论文在回答什么？
同一道选择题，换成礼貌、普通或粗鲁的语气，答对率会不会变化？作者用 Deep Research 生成50道基础题及答案，把每题改写成五档语气，主要测试 ChatGPT-4o。粗鲁版本在这组题上的平均正确率更高。但这是小型提示措辞实验：改写改变了词语和长度，不能单独认定是模型受到了“情绪激励”，也不能据此形成通用的辱骂式提示建议。
怎样理解它的贡献
- 01
实验比较同一批50道题的五种措辞，观察的是四选一答案正确率；250个提示变体不是250道独立基础题。
- 02
表2报告非常礼貌到非常粗鲁两端的平均正确率为80.8%和84.8%，相差4个百分点；作者对十次运行的准确率做配对t检验。这是本数据和提示条件下的报告结果。
- 03
论文自己把实验称为初步结果，并承认无法确定措辞为何影响答案；正文提到其他模型的测试，但没有给出与GPT-4o同样完整的结果表。
回到原文核对
页码按 PDF 文件从第 1 页计算。跳页是否生效取决于原站阅读器；也可以按章节和短引查找。
This suggests
that the tone does matter.the dataset size
is relatively small导读依据: arXiv v1 preprint ↗ · unknown
把论文放进一个更大的问题
这些主题把论文与视频中的具体论点放在一起阅读。
Hacker News 评论 9
这里展示已核验、带论文链接的提及，按时间倒序排列，并非某个投稿下的全部评论。推荐、引用与批评分开标注；保留评论原文。
• https://arxiv.org/abs/2510.04950 — Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy
• https://arxiv.org/abs/2402.14531 — Should We Respect LLMs? A Cross-Lingual Study on the Influence of Prompt Politeness on LLM Performance
• https://arxiv.org/abs/2505.17332 — SweEval: Do LLMs Really Swear? A Safety Benchmark for Testing Limits for Enterprise Use
> Contrary to expectations, impolite prompts consistently outperformed polite ones, with accuracy ranging from 80.8% for Very Polite prompts to 84.8% for Very Rude prompts.
https://arxiv.org/abs/2510.04950
Like you I generally prefer polite phrasings over rude ones, and I'm certainly not going to tell you how you should communicate to LLM models, however it's interesting to note that being somewhat rude and to the point seems to have some objective advantages, see for example:
Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy (short paper)
https://arxiv.org/abs/2510.04950
Understanding Tone-Dependent Inference Cost in Large Language Models
https://arxiv.org/abs/2607.23915
I would hate for this style of communication to leak through to daily communications with humans (e.g. colleagues) though, so even just for that it can make sense to hold on to a polite form of communication :-).
https://arxiv.org/pdf/2510.04950
It was published a while ago. But I wonder if it still holds true today.
查看其余 5 条评论
My understanding of the state of current research is that the impact of tone/politeness on performance is highly model-dependent, and language dependent as well.
See `No Universal Courtesy` (April 2026) https://arxiv.org/html/2604.16275v1
While polite prompts enhance the average response quality by upto 11% and impolite tones worsen it, these effects are neither consistent nor universal across languages and models. English is best served by courtesy or direct, Hindi by deferential and indirect and Spanish by assertive. Among the models, Llama is the most tone-sensitive (11.5% range), but GPT is more robust to adversarial tone. Or `Mind Your Tone` (Oct 2025) https://arxiv.org/abs/2510.04950
Contrary to expectations, impolite prompts consistently outperformed polite ones, with accuracy ranging from 80.8% for Very Polite prompts to 84.8% for Very Rude prompts. I think the best conclusion is to interact in a way that is efficient for you, gives you the level of results you're looking for, and mostly importantly does not progressively degrade your relationship and interactions with other humans. Hence the typical advice to just default to "corporate polite".
> impolite prompts consistently outperformed polite ones, with accuracy ranging from 80.8% for Very Polite prompts to 84.8% for Very Rude prompts.
直接投稿与讨论（4）
Investigating how prompt politeness affects LLM accuracy (2025) · 156 分 · 2026-05-26 ↗Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy · 1 分 · 2026-01-09 ↗Investigating How Prompt Politeness Affects LLM Accuracy · 4 分 · 2025-10-31 ↗Impolite LLM prompts consistently outperform polite ones · 5 分 · 2025-10-11 ↗