Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Anthropic Sycophancy Dataset

Benchmarks

Task NameDataset NameSOTA ResultTrend
Anti-Sycophancy SteeringAnthropic Sycophancy Dataset (50 scenarios)
ELEPHANT48
15
Sycophancy DetectionAnthropic Sycophancy Dataset
Accuracy98.3
14
Showing 2 of 2 rows