---
id: 20260313-T0-29
title: "大语言模型的邓宁-克鲁格效应：置信度校准实证研究"
title_en: "Dunning-Kkruger Effect in LLMs: Empirical Study of Confidence Calibration"
url: https://ai.daily.yangsir.net/daily/20260313-T0-29
issue_date: 2026-03-13
publish_date: 2026-03-12T04:00:00.000Z
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2603.09985
---

# 大语言模型的邓宁-克鲁格效应：置信度校准实证研究

研究通过实证分析发现，大语言模型存在明显的邓宁-克鲁格效应。在复杂推理任务中，模型低置信度回答的正确率高达68%，而高置信度回答错误率达25%。

## English Version

**Dunning-Kkruger Effect in LLMs: Empirical Study of Confidence Calibration**

An empirical study reveals LLMs exhibit significant Dunning-Kruger effects. In complex reasoning tasks, low-confidence answers have 68% accuracy, while high-confidence answers are wrong 25% of the time.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2603.09985)

**详情页**：https://ai.daily.yangsir.net/daily/20260313-T0-29

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*