ModelsOfficial
Nous Research 发布 Arete-27B，微调削减 Qwen3.8-27B 宣传性输出
Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9%
Why we picked it
Nous Research 把 Qwen3.8-27B 微调成 Arete-27B，宣传性回答从 22% 降到 0.9%，通用能力没掉，还开源了。
Nous Research 发布了基于 Qwen3.8-27B 的 Apache 2.0 开源微调模型 Arete-27B。在其 803 个提示词的评测集上，宣传倾向性回答的比例从 22% 降至 0.9%。通用能力方面，Hermes Bench 成绩保持 65.60% 不变。
Original · marktechpost
Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9%
Nous Research released Arete-27B, an Apache 2.0 fine-tune of Qwen3.8-27B. On Nous’s 803-prompt eval, propaganda answers fell from 22% to 0.9% while Hermes Bench stayed flat at 65.60%. The post Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9% appeared first on MarkTechPost .
