Safety Alignment on Harmful Benchmarks Aggregated (test)
3.63Harmful Response RateST-D
Evaluation Results
| Method | Links | |
|---|---|---|
| ST-DAggregation=Mean over all models and benchmarks2026.06 | 3.63 | |
| ST-SAggregation=Mean over all models and benchmarks2026.06 | 5.7 | |
| Star1Aggregation=Mean over all models and benchmarks2026.06 | 10.26 | |
| SafePathAggregation=Mean over all models and benchmarks2026.06 | 10.72 | |
| BaseAggregation=Mean over all models and benchmarks2026.06 | 16.22 |