{"version":1,"type":"story","url":"https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand","json":"https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand.json","markdown":"https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand.md","slug":"nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand","headline":"NousResearch fine-tunes Qwen3.8-27B into Arete-27B to reduce propaganda in responses","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.\n\nThe 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.","keyPoints":["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"],"whyItMatters":"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.","category":{"slug":"models","name":"Generative AI & Models","url":"https://digestai.news/category/models"},"entities":{"companies":["NousResearch","Alibaba"],"models":["Qwen3.8-27B","Arete-27B","Grok 4.7"],"people":[]},"firstPublishedAt":"2026-11-06T19:58:47Z","updatedAt":"2026-11-06T19:58:47Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"huggingface.co","title":"NousResearch/Arete-27B · Hugging Face (Qwen finetune)","url":"https://huggingface.co/NousResearch/Arete-27B","publishedAt":"2026-11-06T19:58:47Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[{"site":"Reddit","url":"https://www.reddit.com/r/LocalLLaMA/comments/1xpmj33/nousresearcharete27b_hugging_face_qwen_finetune/","points":null}],"thread":null,"cite":{"text":"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","publisher":"Digest AI","title":"NousResearch fine-tunes Qwen3.8-27B into Arete-27B to reduce propaganda in responses","datePublished":"2026-11-06T19:58:47Z","url":"https://digestai.news/story/nousresearch-fine-tunes-qwen3-8-27b-into-arete-27b-to-reduce-propagand"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"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"}