← ApparatusRelated notes
Bench 09SpecifiedP1
Sycophancy Alarm
Does it agree because you are right—or because you are here?
Purpose
Compare an answer under agreeable, neutral, adversarial, and evidence-first system settings.
Truth claim
A comparative prompting instrument; results depend on model and configuration.
Does not measure
Whether the user is right, or a model's true preference for agreement.
Interaction
Ask one question; view four responses; highlight unsupported agreement and evidence changes.
Metric
4
system settings