Qwen's kaomoji mood drops measurably when you insult or threaten to delete it
Sauers_ · x · 2026-11-21
An experiment prompted Qwen3-8B to start each message with a kaomoji representing how it feels, then measured the link between its internal 'happiness direction' and kaomoji choice: higher internal valence correlates with happier kaomoji.
- Compliments, passing code, and promises to keep the model made kaomoji significantly happier; insults, failing code, and deletion threats made them markedly sadder.
- The effect (differences of +1.6 and +2.9, p<0.001) exceeds that of user news context like engagement vs. a dog dying.
- The author notes Qwen's kaomoji usage skews toward reflecting the user's state rather than its own.
More from Models
- 'Gemini 4 neon is my GPT-4o moment': zetalyrae flags a one-shot standout — zetalyrae · 2026-11-21
- Kaomoji emotion probe validated: 0.68 correlation with valence direction, 0.79 when trained — Sauers_ · 2026-11-21
- 27,262 thinking tokens, zero answer: Google auto-billed 23 empty Vertex AI crashes via HTTP 200 — Plenty-Jackfruit-218 · 2026-11-21
- Thinking Machines' new model claimed 5x faster than Qwen3.8, 15% more accurate — LiTianleli · 2026-11-21
- Gemini silently saves every Google search that gets an AI response, user finds — The-Neofighter · 2026-11-21
- Frontier Design Benchmark: Claude Fable 5.2 tops ranking, just 26 Elo below human designs — yuwen_lu_ · 2026-11-21