Nous releases Arete-27B to reduce political bias
What happened
On November 7, The Decoder reported that Nous Research fine-tuned Alibaba's Qwen3.8-27B to release Arete-27B, an open-weights model, and analyzed how political bias in Chinese AI models can be reduced through fine-tuning. The results come from Nous Research's own tests: after fine-tuning, the share of propaganda-style answers fell from 22% to 0.9%. What has been disclosed so far is the model release and this test result; the percentages reflect performance in its own tests.
Written by AI from the coverage · updated 12 hours ago
Coverage
Follow the reports to see the story from different sides.
- The DecoderState propaganda in some Chinese AI models is only "a thin layer" that fine-tuning can strip away
Nous Research 基于阿里 Qwen3.8-27B 微调出开放权重模型 Arete-27B，在自建测试中将政治宣传回答比例从 22% 降至 0.9%。
Heat over time
Heat now 7·Comparable peak 10（Nov 7 21:00）·Comparable change over 24 hours –
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.