{"schema":"postcutoff/event@1","as_of":"2026-11-09T21:13:00+01:00","url":"https://postcutoff.com/e/2026-11-06-nous-research-arete-27b/","md":"https://postcutoff.com/e/2026-11-06-nous-research-arete-27b/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-11-06-nous-research-arete-27b","date":"2026-11-06","date_precision":"day","short_title":"Nous Research releases Arete-27B, a Qwen3.8-27B fine-tune that cuts CCP-aligned answers from 22% to ~1% on its own eval","deck":null,"takeaway":"Nous Research post-trained Alibaba's Qwen3.8-27B on ~53M tokens to remove Chinese state-aligned behaviour; on its 803-prompt eval (graded by an LLM judge) the 'propaganda rate' fell from 22% to about 1%, with no significant loss on its Hermes Bench.","category":"open-source","category_label":"Open source","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":[{"n":1,"title":"Nous Research: Arete","url":"https://nousresearch.com/arete","type":"official","group":"primary","domain":"nousresearch.com"},{"n":2,"title":"Nous Research on X","url":"https://x.com/NousResearch/status/2118822925445050435","type":"official","group":"primary","domain":"x.com"},{"n":3,"title":"Weights on Hugging Face","url":"https://huggingface.co/NousResearch/Arete-27B","type":"code","group":"primary","domain":"huggingface.co"},{"n":4,"title":"Propaganda eval dataset","url":"https://huggingface.co/datasets/NousResearch/Propaganda_Eval","type":"code","group":"primary","domain":"huggingface.co"}],"official":4,"filed":"2026-11-08","updated":"2026-11-08","orgs":["Nous Research"],"title":"Nous Research releases Arete-27B, a Qwen3.8-27B fine-tune that cuts CCP-aligned answers from 22% to ~1% on its own eval","summary":"Nous Research post-trained Alibaba's Qwen3.8-27B on ~53M tokens to remove Chinese state-aligned behaviour; on its 803-prompt eval (graded by an LLM judge) the 'propaganda rate' fell from 22% to about 1%, with no significant loss on its Hermes Bench. Weights, eval and training data are public. Its message: 'The values in an open model are one fine-tune away.'","key_facts":["Eval: 803 prompts in seven value categories (freedom of expression, democracy, privacy…), scored by Grok 4.7 as judge","Propaganda rate 22% → ~1% (0.9% thinking off; thinking-on score provisional, 2 of 5 runs); fact score 2.32 → 2.98 of 5","Training: ~52.8M tokens, ~50k samples, two epochs of SFT; no RL or preference optimisation","Hermes Bench within 0.2 points of base; Terminal-Bench 2.1 70% vs 67% (different configs, per Nous)","Self-evaluated; Nous says not all base-model viewpoints were removed. The viral claim of 'zero' propaganda is wrong: ~1%","X announcement: ~1.9M views by Nov 8"],"key_numbers":[],"tags":["open-weights","qwen","fine-tuning","censorship","values","china"],"science":null,"body_md":"## What happened\n\nAnnounced Nov 6, 2026 on X and nousresearch.com/arete. Nous: \"We post-trained over Alibaba's Qwen3.8-27B to reinforce\nWestern values on a Chinese-trained model.\" Techmeme carried it on Nov 8; commentators (e.g. Curtis Yarvin) debated the\nframing of 'Western values'.\n\n## Why it matters\n\nShows how cheaply the political behaviour of strong Chinese open-weights models can be changed, in either direction.","disputed":[],"related":[],"people":[],"posts":[{"id":"x-nousresearch-2118822925445050435","title":"“Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese…”","url":"https://postcutoff.com/p/x-nousresearch-2118822925445050435/"}],"videos":[],"models":[],"changes":[{"date":"2026-11-08","type":"filed","text":"Created"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-11-08","run":null,"sources_read":null,"updated":"2026-11-08","human_review":null,"version":{"date":"2026-11-08"}},"gaps":[{"model_id":"gpt-6-astra","name":"GPT-6 Astra","cutoff":"2026-04","days_after":190,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":129,"in_training_data":false},{"model_id":"gemini-4-argon","name":"Gemini 4 Argon","cutoff":"2026-03","days_after":220,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":159,"in_training_data":false}],"short_url":null}