Unverified50% confidenceSolutionExact time
Prompt design in RAG scenarios has two core requirements: constraining the scope of answers and citing sources, and requiring the model to refuse to answer when context is insufficient.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMeans such as RAG and structured output constraints can alleviate the hallucination problem of the model "misremembering legal provisions" or "fabricating case law," and make conclusions traceable.70% similarUnverified不提供上下文时，本地小模型对领域问题会拒绝回答或产生幻觉，引入 RAG 后能给出准确答案69% similarUnverifiedRAG存在两个漏洞：模型处理上下文时仍可能误读误用，以及检索到的上下文本身可能错误或无关68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1009579API
curl https://kongchang.com/api/v1/knowledge/claims/1009579MCP
get_claim(id=1009579)