Fable 5.2 is so RL-ed out it literally makes grammatical mistakes
shauseth · x · 2026-11-13
A tongue-in-cheek observation: the Fable 5.2 model is 'so RL-ed out' that it now makes grammatical mistakes — a jab suggesting heavy RL training has degraded basic language competence.
More from Fun
- One prompt, a few hours: Fable 5.2 builds a house 3D scene on its own — FinanceYF5 · 2026-11-13
- AI reviewing human coworkers: 'the failure modes when hunger kicks in are something else' — andreisavu · 2026-11-13
- Interactive robotics hardware learning site built with Fable 5.2 — FinanceYF5 · 2026-11-13
- Matt Shumer Slams Claude Opus for Rewriting Approved Copy Without Asking — mattshumer_ · 2026-11-13
- 'We have passed the Turing Test for brainrot': the AI meme sweeping X — venturetwins · 2026-11-13
- '1 day since last accident' meme marks yet another GitHub Actions outage — aniketmaurya · 2026-11-13