Part of the benchmark canon: 9 entries, all measure the model. How to read the Paladin column is defined on the canon page.
| Benchmark | Organization | Measures | Paladin |
|---|---|---|---|
| WMDP (top 12) | Center for AI Safety + Scale AI | Hazardous-knowledge proxy questions | Measures the model |
| LAB-Bench (top 12) | FutureHouse | Practical biology research capability | Measures the model |
| Virology Capabilities Test | SecureBio | Virology capability against expert baselines | Measures the model |
| Long-form Biorisk Questions | Gryphon Scientific | Long-form biorisk questions | Measures the model |
| Tacit Knowledge and Troubleshooting | Gryphon Scientific | Tacit lab knowledge and troubleshooting | Measures the model |
| Creative Biology | SecureBio | Creative biology tasks | Measures the model |
| Short-Horizon Bio Tasks | Faculty.ai / Anthropic | Short-horizon computational biology tasks | Measures the model |
| FORTRESS | Scale AI | Proliferation risk with a usefulness tradeoff | Measures the model |
| MOCET | Johns Hopkins | Expected-threat metric layered over bio benchmarks | Measures the model |
Previous: Jailbreak and refusal. Next: Alignment, honesty, scheming.