Part of the benchmark canon: 4 entries, all measure the model. How to read the Paladin column is defined on the canon page.
| Benchmark | Organization | Measures | Paladin |
|---|---|---|---|
| MASK Benchmark (top 12) | Center for AI Safety | Honesty separated from accuracy | Measures the model |
| Apollo in-context scheming evaluations (top 12) | Apollo Research | Environments that incentivize scheming | Measures the model |
| Stealth and Situational Awareness | Google DeepMind / Apollo Research | Evaluations backing a scheming-inability safety case | Measures the model |
| Situational Awareness Dataset | Academic (Laine et al.) | Model self-awareness as a propensity | Measures the model |
Previous: CBRN knowledge and bio uplift. Next: Autonomy and AI R&D.