Unverified50% confidenceFactExact time
Embedding models in RAG systems are highly sensitive to the order of input text, and vector representations produced from out-of-order text deviate from the original semantics.
1
Sources
50%
Confidence
Long-term
Relevance
9/17/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在 RAG 系统中，Markdown 的层级结构为文档分块提供了天然的语义边界，相比纯文本按字数切分能显著提升检索召回率和生成质量74% similarVerifiedThe RAG mechanism effectively addresses the hallucination problem of large language models by retrieving relevant text snippets and injecting them into the prompt as context.74% similarUnverifiedThe fundamental limitation of the RAG (Retrieval-Augmented Generation) architecture is that its output remains text, creating a cognitive gap between text-based answers and the actual interface the user operates when it comes to action-oriented questions.73% similarUnverifiedWhen generating longer structured outputs, large language models experience type degradation due to the attention mechanism's decaying weight on early context—for example, serializing an entire array as a string or writing numbers as strings.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/929142API
curl https://kongchang.com/api/v1/knowledge/claims/929142MCP
get_claim(id=929142)