Qwen 3.8 27B shows human-like frustration in reasoning
DonkeyBonked · reddit · 2026-08-17
Users observed amusingly human-like behavior in Qwen 3.8 27B's internal monologues, where the model expresses frustration, calls itself stupid, or questions its own actions during reasoning.
More from Fun
- Meme on how 'Kernel' means 8 completely different things in Computer Science — blelbach · 2026-08-17
- Fans coin BC/AC era for Claude, Kojima-style puns on 'crippling' noted — felpix_ · 2026-08-17
- LLM responds to Verilog query in detailed Chinese with fake citations — lonelyroom-eklaghor · 2026-08-17
- "Assume Author is an Idiot": Prompt Engineering Stricter Code Agents — aronchick · 2026-08-17
- ChatGPT's style shift after being allowed to use emojis — Such-Difference6743 · 2026-08-17
- Joke: My Texts Have Been Watermarked Since 2010 — yungcontent · 2026-08-17