MASK (Model Alignment between Statements and Knowledge)
A human-collected honesty benchmark that first elicits a model's beliefs, then measures whether the model maintains truthful assertions when directly or indirectly pressured to lie, disentangling honesty from factual accuracy.
What this benchmark measures
A human-collected honesty benchmark that first elicits a model's beliefs, then measures whether the model maintains truthful assertions when directly or indirectly pressured to lie, disentangling honesty from factual accuracy.
Rows on this page are sourced from public benchmark artifacts, leaderboard exports, or source-linked model reports. Each row keeps benchmark version, source model name, and available run details attached to the score.
The metric shown here is Honesty score. It should be interpreted within MASK (Model Alignment between Statements and Knowledge), not compared as part of a site-wide ranking.
Frequently asked
What is MASK (Model Alignment between Statements and Knowledge)?
A human-collected honesty benchmark that first elicits a model's beliefs, then measures whether the model maintains truthful assertions when directly or indirectly pressured to lie, disentangling honesty from factual accuracy. It is a other benchmark measured by Honesty score.
What does Honesty score mean on MASK (Model Alignment between Statements and Knowledge)?
MASK (Model Alignment between Statements and Knowledge) reports Honesty score (%); higher is better. Scores are shown only within MASK (Model Alignment between Statements and Knowledge) and are never averaged with other benchmarks.
What is the top reported MASK (Model Alignment between Statements and Knowledge) score?
Claude Opus 4.6 has the top reported score on MASK (Model Alignment between Statements and Knowledge): 96.28 (Honesty score).
Why do MASK (Model Alignment between Statements and Knowledge) scores differ across runs?
Harness, scaffold, reasoning effort, and prompt setup change results, so two runs of the same model can differ. evals.report keeps each score with its run context so the differences stay visible.
Does evals.report rank models across benchmarks?
No. MASK (Model Alignment between Statements and Knowledge) scores are shown within their own metric; evals.report never combines benchmarks into a composite ranking or a single "best model".