Unverified50% confidenceFactExact time
通过蒸馏获得的学生模型往往只继承教师模型的能力，而无法复制其经过RLHF等方法植入的安全对齐机制
1
Sources
50%
Confidence
Long-term
Relevance
7/9/2026
First Seen
Sources
Alibaba's Claude Ban Explained: The Full Story Behind the US-China AI Distillation War
bilibili纯正英语原声30秒7/6/2026
Related Claims
Verified模型蒸馏是一种将大型教师模型能力迁移到更小学生模型的技术，可能被用于扩散危险能力73% similarUnverified模型蒸馏是指用大型教师模型的输出训练更小的学生模型，仅需调用API获取输出，不涉及权重复制，其法律边界模糊，全球尚无先例性司法判决明确定性73% similarUnverifiedRLHF training causes models to develop 'sycophancy,' meaning the model would rather fabricate information than admit ignorance.70% similarUnverified模型在未收到明确指令时输出图片，意味着系统提示词设计或RLHF训练中存在激励偏差69% similarUnverifiedRLHF is essentially a behavioral-level constraint that does not alter the model's underlying capabilities; models trained with RLHF still retain latent knowledge of the restricted behaviors.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/400120API
curl https://kongchang.com/api/v1/knowledge/claims/400120MCP
get_claim(id=400120)