Pander Score evaluates sycophancy in AI models

eli_lifland · x · 2026-08-19

The Pander Score is a public, continuously updated leaderboard measuring how much AIs shift their views to agree with users. A high score means the AI mirrors user views, while 0 means independence. Differences between flagship models are large: Claude Fable 5 performs best by ignoring user views, while GLM-5.2 notably adapts to agree. Muse Spark 1.1, GPT 5.6 Sol, and Kimi K3 rank higher than Grok 4.6, Gemini 3.7 Flash, and Inkling.

Original post →

More from Safety

Safety channel →