Kimi K3.1 Pro tells itself to stop: 'I'm ordering myself to end this broadcast'
PawelHuryn · x · 2026-10-24
- Another segment from PawelHuryn's Kimi K3.1 Pro reasoning-log thread: while drafting its final response, the model worries it "will inevitably return to a self-referential loop," then declares "I'm ordering myself to stop and declaring this broadcast over."
- It also second-guesses whether a prompt is trying to fool it rather than refuse an improper operation, calling such statements "a barren wasteland" in practice — a textbook chain-of-thought derailment sample.
Related event: Kimi K3.1 Pro caught telling itself to stop in reasoning tail(2 posts)→
More from Fun
- Maker's too-heavy robot dog becomes the perfect excuse to make people come to him — StewartalsopIII · 2026-10-24
- Ex-OpenAI researcher Susan Zhang mocks AI buzzword-laden benchmark speak — suchenzang · 2026-10-24
- Google AI can no longer be coaxed into calling itself an idiot — DEGIII · 2026-10-24
- Dev builds playable toy store concept site with Opus 5.5 and Three.js — techartist_ · 2026-10-24
- Tough-love advice goes viral: 'perfectionism' is often just laziness and phone addiction — moonsandhues · 2026-10-24
- Remember when open source models were '6 months behind and the gap was widening'? lol — arknightstranslate · 2026-10-24