DeepSeek蒸馏论的一个反例：前沿实验室故意压制的任务
有人质疑 DeepSeek 全靠蒸馏，这条推文提出一个挺刁钻的反驳角度：在实验室故意压分的任务上提分，蒸馏解释不通。
文章讨论了围绕 DeepSeek 的一个争议：其性能提升是否全部来自对前沿模型的蒸馏。作者指出一个关键细节：前沿实验室在某些任务上会故意压制模型表现，即所谓的 sandbagging。
如果 DeepSeek 在这些任务上仍然取得提升，就难以用蒸馏解释。
One interesting wrinkle in the "all of DS gains are down to distillation" will be gains on things that frontier labs deliberately sandbag on.
This report is categorized under AITOP's "Models" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI regulation, compliance frameworks, and sovereign compute policy. Key entities and technical topics under continuous tracking include DeepSeek, 蒸馏, 推理模型, AI安全.
Source attribution: curated from Teortaxes (@teortaxesTex).
