Unverified50% confidenceTradeoffExact time
RAG系统的检索模块主要依赖语义相关度而非信源可信度筛选文档，导致语义相关的假新闻可能比语义稍弱的权威报道获得更高检索优先级
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
AI谣言如何"杀死"金·凯瑞：搜索时代的信任危机
hackernewshackernews7/4/2026
Related Claims
UnverifiedRAG系统的效果高度依赖于知识库质量，源文档提取阶段的信息丢失会受到垃圾进垃圾出的制约74% similarUnverifiedRAG（检索增强生成）、事实核查链和置信度评分是对抗LLM幻觉的多层防御机制73% similarUnverifiedHigh-authority information sources dominate retrieval rankings in RAG systems while long-tail content rarely wins the competition to enter the final context window73% similarVerifiedRAG（检索增强生成）在调用LLM前从向量数据库检索相关文档片段注入提示词，当检索文档不匹配或包含错误信息时会产生'由检索驱动的幻觉'73% similarUnverified当前RAG系统的检索模块主要依赖语义相关度而非信源可信度筛选文档，这是其在事实核查场景下的核心架构性缺陷73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/198744API
curl https://kongchang.com/api/v1/knowledge/claims/198744MCP
get_claim(id=198744)