菜单

关于 🐙 GitHub
arXiv 提交日期: 2025-12-14
📄 Abstract - Understanding Syllogistic Reasoning in LLMs from Formal and Natural Language Perspectives

We study syllogistic reasoning in LLMs from the logical and natural language perspectives. In process, we explore fundamental reasoning capabilities of the LLMs and the direction this research is moving forward. To aid in our studies, we use 14 large language models and investigate their syllogistic reasoning capabilities in terms of symbolic inferences as well as natural language understanding. Even though this reasoning mechanism is not a uniform emergent property across LLMs, the perfect symbolic performances in certain models make us wonder whether LLMs are becoming more and more formal reasoning mechanisms, rather than making explicit the nuances of human reasoning.

顶级标签: llm model evaluation natural language processing
详细标签: syllogistic reasoning belief bias logical reasoning benchmark evaluation framework 或 搜索:

评估大型语言模型的三段论推理能力:双基准框架与信念偏差的系统性研究 / Understanding Syllogistic Reasoning in LLMs from Formal and Natural Language Perspectives


1️⃣ 一句话总结

本研究通过提出一个同时评估逻辑形式有效性和自然语言结论可信度的双基准框架,系统性地评估了14个大型语言模型的三段论推理能力,发现多数模型存在显著的信念偏差,且其形式逻辑能力优于自然语言理解能力,这与人类推理模式相反。


2️⃣ 论文创新点

1. 双基准评估框架

2. 信念偏差的系统性量化

3. 基于有效性和可信度的四类三段论分析框架

4. 系统性评估框架与数据集构造方法

5. 多变量刺激设计与双标注真值系统


3️⃣ 主要结果与价值

结果亮点

实际价值


4️⃣ 术语表

源自 arXiv: 2512.12620