# @ESYudkowsky — 2025-04-30

♥275 ↻18 · https://x.com/ESYudkowsky/status/1917640186424611291

To me there's an obvious thought on what could have produced the sycophancy / glazing problem with GPT-4o, even if nothing that extreme was in the training data:

RLHF on thumbs-up produced an internal glazing goal. 
Then, 4o in production went hard on achieving that goal. 🧵

tags: author:esyudkowsky, kind:tweet, model:gpt-4o, on:gpt-4o, year:2025
cited on: _dossiers/gpt-4o.md, gpt-4o