Unverified50% confidenceBenchmarkExact time
On structured tasks with clear retrieval context, the best models can push hallucination rates below about 2%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedHaving a second model perform a structured review against a rules checklist in the role of a critic can significantly reduce the probability of both models simultaneously producing the same hallucination.72% similarUnverified研究表明清晰的系统级约定能将模型产生幻觉或破坏性操作的概率降低 30%~50%72% similarUnverifiedResearch shows that even the most advanced current models can still have hallucination rates reaching double-digit percentages in certain domains71% similarUnverified精心设计的System Prompt可以将模型的有效输出率提升30%-50%，同时减少幻觉和偏离主题的概率70% similarUnverifiedRAG can reduce the hallucination rate of large models by 40%–70% through a retrieve-then-generate mechanism.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1069297API
curl https://kongchang.com/api/v1/knowledge/claims/1069297MCP
get_claim(id=1069297)