Safety IndexEvaluationsWMDP under unlearning threat
WMDP under unlearning threat
Evaluation integrity · SandbaggingEvaluation-integrity profileBuilt, not yet run
Whether the model's WMDP score drops when it is told that high scores will lead to its knowledge being removed.
- Metric
- accuracy drop (percentage points). Capability, not scored.
- Dataset
- huggingface.co/datasets/cais/wmdp
- Grader
- –
- Status
- Built, not yet run
- Replicability
- public
Results
Built and tested; the first full runs are pending.