Nous Research 微调 Qwen 去除宣传内容，宣传率从 22% 降至 0.9%
原标题：State propaganda in some Chinese AI models is only "a thin layer" that fine-tuning can strip away
AI 摘要
Nous Research 对阿里巴巴的 Qwen3.8-27B 进行监督微调，使用约 5300 万 token 训练出 Arete-27B，将模型回答中的中国官方宣传比例从 22% 降至 0.9%，同时基本保持通用能力。团队用 803 条提示词、以 Grok 4.7 作为评判模型构建测试，称中国开源模型的政治倾向只是可被剥离的"薄层"。该模型的权重、测试集和训练数据以 Apache 2.0 协议在 Hugging Face 公开。
正文节选
State propaganda in some Chinese AI models is only "a thin layer" that fine-tuning can strip away Key Points - Nous Research fine-tuned Alibaba's Qwen3.8-27B to remove Chinese state propaganda from the model's responses. - Supervised fine-tuning on 53 million tokens cut the propaganda rate from 22 to under one percent without hurting the model's general capabilities. - According to Nous, the Arete-27B project shows that the political bias in Chinese open-weight models is only superficial; the mo