AI researchinfo
Nous Research Releases Arete-27B Fine-Tuned Qwen Model
Original title: Nous Research Releases Arete-27B that Cuts Qwen3.8-27B Propaganda From 22% to 0.9%
AI summary · x-ai/grok-4.7Summaries by Vani AI
Arete-27B is a 27B open-weight model reducing propaganda answers from 22% to 0.9% on an 803-prompt eval.
Nous Research released Arete-27B, a supervised fine-tune of Alibaba's Qwen3.8-27B. The 27B dense parameter model strips state-aligned answers while keeping core capabilities. Weights, training data, and evaluation set are open under Apache 2.0.
Key facts
- Propaganda rate fell from 22% to 0.9% on 803-prompt PropagandaEval.
- Fact score rose from 2.32 to 2.98.
- Trained on 52.8M tokens using only SFT.
- BF16 weights are 53.81 GB; Q4KM quant is 17.20 GB.
- License matches base model at Apache 2.0.
- Vendors
- Nous ResearchAlibaba
- AI models
- Arete-27BQwen3.8-27B
From the original article
7 min read at marktechpost.comNous Research released Arete-27B, an Apache 2. 0 fine-tune of Qwen3. 8-27B.
By Asif Razzaq · marktechpost.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.