Expired60% confidenceFactExact time
据斯坦福大学2024年的研究，即使是最先进的模型，在事实性任务中的幻觉率仍然在3%-15%之间波动
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
5/27/2026
First Seen
Valid until: 8/25/2026(expired)
Sources
Thinking Can Be Outsourced, Understanding Cannot: The Most Overlooked Cognitive Blind Spot of the AI Era
twitterOfficialLoganK5/27/2026
Related Entities
Related Claims
VerifiedResearch shows that even the most advanced models have hallucination rates in factual question-answering tasks that still fluctuate between 5% and 20%.80% similarUnverified早期模型在某些领域的事实幻觉率高达15-25%，尤其在长尾知识领域幻觉率更高76% similarUnverifiedA 2023 Stanford University study found that even the most advanced models at the time experienced a 30% to 50% drop in information-retrieval accuracy for middle segments when processing contexts exceeding 32K tokens.72% similarUnverifiedStanford research shows that in complex debugging scenarios, the probability of LLM suggestions causing beginners to form incorrect mental models is approximately 23%, far higher than that of human tutors.71% similarUnverifiedA study from Vectara shows that even the most advanced RAG systems can still have hallucination rates of 3%–10%.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/35947API
curl https://kongchang.com/api/v1/knowledge/claims/35947MCP
get_claim(id=35947)