---
id: 20260702-T0-04
title: "CORTEX：通过对比内部表示实现 RAG Token 级幻觉检测"
title_en: "CORTEX: Token-Level Hallucination Detection in RAG Systems"
url: https://ai.daily.yangsir.net/daily/20260702-T0-04
issue_date: 2026-07-02
publish_date: 2026-07-01T04:00:00.000Z
category: research
source_name: "arXiv cs.CL (NLP)"
source_url: https://arxiv.org/abs/2606.31033
---

# CORTEX：通过对比内部表示实现 RAG Token 级幻觉检测

针对检索增强生成（RAG）中的幻觉问题，新论文提出了 CORTEX 检测法。研究发现长文本 RAG 输出中的幻觉通常呈现局部化特征，并非全篇失效。CORTEX 通过对比模型的内部表示，能够在 Token 级别上精准定位幻觉片段，为提升 RAG 可靠性提供了新工具。

## English Version

**CORTEX: Token-Level Hallucination Detection in RAG Systems**

A new paper proposes CORTEX, a method for detecting hallucinations at the token level in Retrieval-Augmented Generation (RAG). Since hallucinations in long-form outputs are often localized, CORTEX uses comparative internal representations to identify specific spans of incoherent text, offering a practical tool for improving RAG reliability.

---

**来源**：[arXiv cs.CL (NLP)](https://arxiv.org/abs/2606.31033)

**详情页**：https://ai.daily.yangsir.net/daily/20260702-T0-04

---

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