# Alibaba’s Qwen3.8 re-educated with a more Western worldview

- URL: https://talkfuse.com/story/alibabas-qwen3-8-re-educated-with-a-more-western-worldview/
- Published: 2026-11-13T01:25:00+00:00
- Updated: 2026-11-13T03:03:58+00:00
- Original publisher: The Register
- Original article: https://www.theregister.com/ai-and-ml/2026/11/13/alibabas-qwen38-re-educated-with-a-more-western-worldview/5309555
- Topics: Asia, World

## Key points

- Nous retrained Alibaba’s Qwen3.8-27B as Arete-27B.
- It says propaganda-classified answers fell from 22% to 1%.
- The work used 50 million tokens and Inkling as a teaching model.

## Summary

Nous Research, an American AI lab, says it has retrained Alibaba’s open-weight Qwen3.8-27B model to reduce Chinese state-aligned responses, releasing the result as Arete-27B. The work highlights how training can shape a model’s answers on politically sensitive topics, and raises questions about how claims of “Western values” are defined.

Nous reports that Arete’s fact score rose from 2.32 to 2.98 out of five, while answers it classified as propaganda fell from 22 per cent to 1 per cent, after training on 50 million tokens. It used the US-made open-weight model Inkling as a teaching model; separately, Hirundo reported reducing similar outputs in another Qwen model from 89.8 per cent to 2.8 per cent. These are company-reported evaluations, and the article notes broader concerns about model bias and the use of AI outputs to train competing systems.

---
Summarised by TalkFuse (https://talkfuse.com/). Quote the original publisher for the reporting and link this page for the summary and discussion.