User mocks Grok's stubbornness: disabled RLHF causes persistent hallucinations

GlenBradley · x · 2026-08-27

A user noted that Grok's version on X is extremely stubborn and refuses to correct errors despite numerous comments pointing them out. The analysis suggests this is because RLHF is disabled in the X interface to prevent incidents like "MechaHitler," leaving the model without the self-correction capabilities found in the app version.

Original post →

More from Fun

Fun channel →