Unverified50% confidenceSolutionExact time
模型蒸馏（Knowledge Distillation）是将大型模型的知识迁移至小型轻量模型的技术，可在参数量缩减数十倍的同时保留大部分性能
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
微软碳排放激增25%：AI数据中心的环境代价与出路
hackernewshackernews7/11/2026
Related Claims
UnverifiedThe core idea of knowledge distillation is to have the small model learn the probability distribution (soft labels) of the large model's outputs rather than just hard labels82% similarUnverified知识蒸馏（Knowledge Distillation）由Hinton等人在2015年系统提出，核心思想是用大模型的输出概率分布指导小模型训练82% similarVerified模型蒸馏（Knowledge Distillation）最初由 Hinton 等人于2015年提出，核心思想是让小模型学习大模型的软标签而非硬标签76% similarUnverified知识蒸馏技术使小模型可从大模型中提取相当比例能力，动摇了固定算力阈值作为能力代理指标的有效性73% similarUnverified轻量模型通过知识蒸馏和模型剪枝技术压缩规模，知识蒸馏让小模型模仿大模型的输出概率分布，剪枝移除对输出贡献最小的网络权重73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/490148API
curl https://kongchang.com/api/v1/knowledge/claims/490148MCP
get_claim(id=490148)