What a model looks like after you RL-fry it
burny_tech · x · 2026-10-08
burnytech shares a demo image of what happens when a model is over-trained with RL, captioned "when you very RLfry your model" — a look at the weird, degenerate behavior that emerges from excessive reinforcement learning fine-tuning.
More from Fun
- Apple gave us 120 acres: employee recalls Apple Park's campus perks — jdluk87 · 2026-10-08
- Watching my AI agent do my job while I just say 'looks good' and 'continue' — tekbog · 2026-10-08
- Vibecoding a racing game: players build their own cars with Claude and compete — Daniel_Farinax · 2026-10-08
- Elon Musk's 1984 game Blastar remade with Grok, now playable inside X — Daniel_Farinax · 2026-10-08
- Mocked as 'AI psychosis', the claim turns out to be a real new planet discovery — rickasaurus · 2026-10-08
- ChatGPT solves user's Linux bug by digging up his own old Reddit post — Melted_VanillaStick · 2026-10-08