📄 中文摘要
通过对Claude（Sonnet）模型进行实验，将同一问题同时发送给两个实例，其中一个使用标准提示，另一个在系统提示中嵌入了来自论文的3个数学公理。结果显示，增强版本返回的答案在结构上更为深刻，且并未提及公理或使用任何术语，深度自然显现。为了方便他人进行比较，作者构建了一个Claude工具，任何人都可以尝试这一方法。
我在Claude的系统提示中添加了3个公理，推理质量显著改变——这是并排比较工具
📄 English Summary
I Added 3 Axioms to Claude's System Prompt and the Reasoning Quality Visibly Changed — Here's the Side-by-Side Comparison Tool
An experiment was conducted using Claude (Sonnet) by sending the same question simultaneously to two instances: one with a standard prompt and the other with three mathematical axioms embedded in its system prompt. The enhanced version produced structurally deeper answers without mentioning the axioms or using any jargon, allowing the depth to appear naturally. To facilitate further exploration, a comparison tool was created, enabling anyone to try this approach.