English(EN) Detecting Hallucinations in Retrieval-Augmented Generation through Grounding-Aware Sensitivity by Perturbation (GASP)

新的GASP方法可检测RAG系统中的句子级幻觉

作者 PulseAugur 编辑部 · [2 个来源] · 2026-07-05 10:35

研究人员开发了一种名为扰动感知基础的敏感性（GASP）的新方法，用于检测检索增强生成（RAG）系统中的幻觉。与提供单一分数的先前方法不同，GASP能够识别答案中未被检索到的证据支持的特定句子。该技术衡量当支持性上下文被移除时，句子可能性的变化程度，从而区分基础内容和幻觉内容。 AI

影响该方法通过对不受支持的陈述提供更细粒度的反馈，有助于开发更值得信赖的AI，从而提高RAG系统的可靠性。

排序理由该集群包含一篇学术论文，详细介绍了检测AI幻觉的新方法。

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.AI TIER_1 English(EN) · Mohamed Aly Bouke · 2026-07-07 04:00

Detecting Hallucinations in Retrieval-Augmented Generation through Grounding-Aware Sensitivity by Perturbation (GASP)

arXiv:2607.04223v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) reduces but does not eliminate hallucination, and existing detectors return a single answer-level score that does not indicate which sentence is unsupported, or why. To close this gap, we intro…
arXiv cs.CL TIER_1 English(EN) · Mohamed Aly Bouke · 2026-07-05 10:35

通过基于地面感知的扰动敏感性（GASP）检测检索增强生成中的幻觉

Retrieval-augmented generation (RAG) reduces but does not eliminate hallucination, and existing detectors return a single answer-level score that does not indicate which sentence is unsupported, or why. To close this gap, we introduce Grounding-Aware Sensitivity by Perturbation (…