Unverified60% confidenceFactExact time
The distribution of faces across different ethnicities is extremely imbalanced in mainstream text-to-image model training datasets (such as LAION-5B), with white faces comprising the overwhelming majority
2
Sources
60%
Confidence
Long-term
Relevance
7/17/2026
First Seen
Sources
Related Claims
UnverifiedNIST's 2019 tests found that certain facial recognition algorithms had false identification rates for Black faces that were 10 to 100 times higher than for white faces.67% similarUnverifiedFacial recognition systems have significantly higher misidentification rates for people with darker skin than for those with lighter skin.63% similarUnverified南亚裔和黑人面孔在多数文生图模型中的还原度较差，模型容易过拟合退化成失真形象60% similarUnverified面部年龄估算技术在不同种族群体中的准确率存在显著差异59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/541283API
curl https://kongchang.com/api/v1/knowledge/claims/541283MCP
get_claim(id=541283)