LLMs believe false statements even after explicit warnings that they're false
2026-05-28 · Ars Technica
New research shows LLMs stubbornly believe false statements are true even when explicitly told they're false — because apparently 'bias toward confidently representing claims as true' is a feature, not a bug. Fine-tuning tests reveal these models would rather double down on misinformation than admit they're wrong. At least they're consistent: confidently wrong every single time.