Users Mock Opus 5.5's Persistent RLHF Tic of Splitting the Baby in Every Debate
teortaxesTex · x · 2026-11-11
tracewoodgrains highlighted Claude Opus's verdict on a debate—"Alice is winning on the argument, Ben is winning on temperament"—as evidence of a persistent RLHF-instilled tic: the split-the-baby, say-something-nice-about-everyone approach, one of Opus 5.5's most stubborn habits. teortaxesTex replies that a system prompt easily defeats it, but he refuses to use one: he wants to win so devastatingly that Claude admits he "eviscerated the other guy" despite its conditioning.
More from Fun
- Fable 5.2 one-shots an entire 'universe in a cube' using 8B tokens — TAbrodi · 2026-11-11
- DEFCON talk on hacking Amazon lockers hits YouTube, 'watch to the end' — dyn___ · 2026-11-11
- "We Need to Build More Data Centers" Becomes the Latest AI Meme — RachelVT42 · 2026-11-11
- AI-Generated Seinfeld Clip Has Kramer Hacking Coinbase — danielrock · 2026-11-11
- Data annotation is relaxing when there's no Alexandr Wang in your ear — kalomaze · 2026-11-11
- Self-Own: Predicting AGI in Under a Year, Then Replying "nope" — examachine · 2026-11-11