Unverified50% confidenceOpinionExact time
Open-weight models may carry hidden biases (such as sexism, political leanings, and cultural differences) due to unknown training data.
1
Sources
50%
Confidence
Long-term
Relevance
9/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOpen-source models may have biases and risks embedded from training data, only the attribution of responsibility for those risks becomes more diffuse80% similarUnverifiedOpenAI在InstructGPT论文中证明，少量高质量人类偏好数据对模型对齐的贡献远超大规模无监督训练75% similarUnverifiedIf training data contains historical biases against specific ethnicities, income classes, or regions, the model will entrench and scale those biases.73% similarUnverifiedResearch from institutions including OpenAI and Anthropic indicates that models trained with RLHF systematically bias toward producing answers that satisfy raters rather than answers that are objectively accurate.72% similarUnverifiedMultiple mathematicians have publicly questioned whether OpenAI models' training data includes unauthorized or even unpublished academic research.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/940245API
curl https://kongchang.com/api/v1/knowledge/claims/940245MCP
get_claim(id=940245)