●builderYou can use AEDI to benchmark your deployed model's susceptibility to user-attitude-driven answer drift before shipping to high-stakes users.
●researcherAEDI's continuous scoring protocol and the LLM-as-judge probability estimation method offer a reusable evaluation framework for sycophancy measurement.
●policyA standardized continuous sycophancy metric across prominent models gives regulators and auditors a concrete tool for epistemic reliability assessment.