Unverified50% confidenceBenchmarkExact time
2023年多项研究表明，GPT-4在回答医学问题时虚构参考文献的比例可高达30%-70%
1
Sources
50%
Confidence
Long-term
Relevance
8/13/2026
First Seen
Sources
Related Claims
UnverifiedA 2023 study found that GPT-4 produced errors in approximately 15–20% of cases when generating code documentation.75% similarVerifiedEven top-tier models like GPT-4 can have hallucination rates of 15%–30% in specific domains.74% similarUnverified2024 年纽约大学的一项研究发现，GPT-4 生成的代码中约有 5.2% 的包引用指向不存在的软件包73% similarUnverifiedGoogle DeepMind's FactScore shows that GPT-4's proportion of accurate factual statements in biographical generation tasks is approximately 73%.71% similarUnverifiedIn 2023, GPT-4 achieved a passing score of 60 or above on all three stages of the USMLE, with some studies reporting scores approaching the average medical student level.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/742278API
curl https://kongchang.com/api/v1/knowledge/claims/742278MCP
get_claim(id=742278)