ishenwei/nexus
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Auto-sync

Browse Source
This commit is contained in:
weishen
2026-04-15 16:33:26 +08:00
parent d3e7fcf81f
commit 22c7a6e4d9
80 changed files with 1473 additions and 240 deletions
Download Patch File Download Diff File Expand all files Collapse all files

30
wiki/concepts/Source-Grounding.md
View File

@@ -1,30 +0,0 @@
---
id: Source-Grounding
title: "Source-Grounding"
type: concept
tags: [AI, RAG, 知识管理]
sources: []
last_updated: 2026-04-15
---
# Source-Grounding
## Definition
Source-Grounding 是一种 AI 回答质量控制机制，严格将 AI 知识库限制为用户上传的文档，确保每个回答都直接溯源至原文，拒绝生成不可验证的推测性内容。
## Description
NotebookLM 的核心机制。与传统 RAG（Retrieval-Augmented Generation）不同，Source-Grounding 不依赖外部知识库，而是以用户提供的文档作为唯一事实来源，从根本上消除幻觉（Hallucination）问题。
## Key Properties
1. **知识边界清晰**：AI 仅能访问上传文档，无法访问外部信息
2. **引文强制**：每个回答必须附带原文引文，可点击跳转验证
3. **准确性优先**：以牺牲通用性换取可信度，适用于法律、医学、项目文档等高精度场景
## Related Concepts
- [[AI知识库]]：依赖外部知识检索，存在幻觉风险
- [[RAG]]：检索增强生成，但依赖广泛语料库
- [[被动学习]]：Source-Grounding 使被动学习成为可能
## Connections
- [[NotebookLM]] ← 实现方 ← [[Source-Grounding]]
- [[引文追溯]] ← 关键特性 ← [[Source-Grounding]]