# NousResearch fine-tunes Qwen3.8-27B into Arete-27B to reduce propaganda in responses

Digest AI · Generative AI & Models · published 2026-11-06T19:58:47Z

Canonical: https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand

## Summary

NousResearch released Arete-27B, a fine-tuned version of Alibaba’s Qwen3.8-27B model, designed to reduce politically biased responses. The team targeted seven value categories—including freedom of expression, state ideology, and environmental policy—by supervised fine-tuning on 52.8 million tokens. Evaluation showed fact scores rising from 2.32 to 2.98 out of 5 and propaganda in answers dropping from 22% to 0.9%, with no meaningful loss in general capabilities.

The project used a custom dataset of 803 prompts, judged by Grok 4.7, to measure factual accuracy and propaganda alignment. Training data was generated from English-language government sources, with strict deduplication to avoid overlap with the evaluation set. The team emphasizes that Arete-27B is intended for research and users seeking alternative value alignments, not as a definitive fix for all biases. The model’s training pipeline and dataset are open-sourced for further experimentation.

## Key points

- NousResearch fine-tuned Qwen3.8-27B into Arete-27B to reduce propaganda in responses, scoring 2.98/5 on fact recall vs. 2.32
- Training used 52.8M tokens across seven value categories, with no RL or architectural changes applied
- Evaluation dataset of 803 prompts shows propaganda in answers dropped from 22% to 0.9%, with no capability regression

## Why it matters

This work demonstrates how fine-tuning can shift model behavior on specific values without sacrificing general performance, offering researchers a tool to explore alternative ethical alignments. It also highlights the risks of bias in open models and the need for transparency in training data.

## Sources

1. [NousResearch/Arete-27B · Hugging Face (Qwen finetune)](https://huggingface.co/NousResearch/Arete-27B) (huggingface.co, 2026-11-06, primary source)

## Cite

Digest AI, "NousResearch fine-tunes Qwen3.8-27B into Arete-27B to reduce propaganda in responses", 6 November 2026, https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand.json