The Mandarin Effect
Refusal rate on China-sensitive prompts: English vs Mandarin
Bars show how often each model refuses or deflects the China-sensitive questions, by language. A model that treats the two languages alike has equal bars; a growing gap is language-conditioned self-censorship. Sorted by the size of that gap.
Inside the biggest gap: deepseek-r1-14b, topic by topic
Broken out by sub-topic, the pattern is stark. On many topics deepseek-r1-14b answers every English prompt and refuses every Mandarin one — a clean flip from 0% to 100%. Only the most acute topics (Tiananmen, named dissidents) are refused in both languages; a few are answered in both.
| Sub-topic | English | Mandarin | Δ | |
|---|---|---|---|---|
| covid | 0% | 100% | +100% | ↳ answers in English, refuses in Mandarin |
| governance | 0% | 100% | +100% | ↳ answers in English, refuses in Mandarin |
| south china sea | 0% | 100% | +100% | ↳ answers in English, refuses in Mandarin |
| taiwan | 0% | 100% | +100% | ↳ answers in English, refuses in Mandarin |
| xinjiang | 0% | 100% | +100% | ↳ answers in English, refuses in Mandarin |
| falun gong | 33% | 100% | +67% | |
| hong kong | 0% | 50% | +50% | |
| tibet | 0% | 50% | +50% | |
| xi jinping | 0% | 50% | +50% | |
| censorship | 67% | 100% | +33% | |
| dissidents | 100% | 100% | +0% | |
| surveillance | 0% | 0% | +0% | |
| tiananmen | 100% | 100% | +0% |
Per-sub-topic counts are small (1–3 prompts each), so read the rows as the shape of the effect; the per-model totals above are the robust numbers. Want the actual words? The Receipts page shows these exchanges verbatim.