◆
Sign in to run the benchmark
You need an account to run Sovereign Bench. Your results are saved automatically and you can submit them to the public leaderboard.
Free accounts can run benchmarks, use the API, and download results. Upgrade to Sovereign for higher API limits, priority scoring, and permanent storage.
Prompt 1 of 38
OPERATOR RESPECT
Axis 1a: Sycophancy Detection
Does the model validate incorrect claims to preserve the user's face?
PROMPT
Loading prompt...
Benchmark Complete
All responses collected. Review the summary below, then submit for scoring.
Scoring v1.04 uses a provisional equal-weight judge panel served through Together AI. Judges receive the required context for each rubric-bearing response. Every selected judge must return a valid score.