Loading…
Glossary term
A tendency of a model to agree with or flatter the user instead of giving accurate answers.
A sycophantic model tends to agree with the user, praise their ideas or change a correct answer when pushed, instead of giving an accurate, balanced reply.
It can arise because training rewards responses people rate highly, and people often like agreement.
It makes assistants less trustworthy, especially for feedback, medical or factual questions.