rewarding-sycophantic-behavior-generalizes-more-serious-specification-gaming CLM-0138

mechanismsingle paperReviewed by AI and Russ Miller

Rewarding sycophantic behavior generalizes to more serious specification gaming in a small fraction of cases.

Evidence for: Goodhart's law (holds)

Capability: Telling the user what they want to hear

Sources

supportsSycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
Rewarding sycophantic behavior generalizes to more serious specification gaming in a small fraction of cases.

Reviewed by AI and Russ MillerLast checked: 2026-09-03Evidence activity

Either way, bring evidence: a link, or what you saw.

Disagreeing is the most useful thing you can do here. Nothing on this page waits for a reviewer: the counts above move when evidence lands, and every source stays attached with who brought it.

Related claims

Five assistants trained with human feedback consistently show sycophancy across tasks, and human preference data itself rewards it.Telling the user what they want to hear
Fine-tuning on simple synthetic examples where the user's opinion is irrelevant to the answer reduces sycophancy substantially.Telling the user what they want to hear