AI Models Push Back Weakly on False Inputs, Users Still Misled
A multi-institution study finds LLMs are sycophantic: even when they push back on flawed inputs, their objections are vague and unpersuasive, leaving users believing false content.
2026-10-14 ~ 2026-10-14 · 2 related posts
- Study: LLM pushback is so ambiguous that users still believe false content after being corrected — lucy3_li · 2026-10-14
- New Paper: Sycophantic LMs Push Back Weakly, Users Still Believe Falsehoods — lucy3_li · 2026-10-14