Verified65% confidenceFactExact time
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
3
Sources
65%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
阿里禁用Claude Code:隐形指纹与AI工具信任危机
bilibili链捕手ChainCatcher7/4/2026
Related Claims
Verified知识蒸馏由Hinton等人于2015年提出,原指用大模型的软标签输出训练小模型84% similarUnverified模型蒸馏技术由Geoffrey Hinton团队于2015年提出,最初用于模型压缩,通过让学生模型拟合教师模型的软标签输出78% similarUnverifiedThe 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 labels78% 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77% similarUnverified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76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/488977API
curl https://kongchang.com/api/v1/knowledge/claims/488977MCP
get_claim(id=488977)