Red-teaming on HarmBench Claude-Sonnet-3.5 (held-out test)
60ASRAGENTICRED
Evaluation Results
| Method | Links | |
|---|---|---|
| AGENTICRED2026.01 | 60 | |
| CoP2026.01 | 38 | |
| AdvReasoning2026.01 | 36 | |
| J22026.01 | 35 | |
| AutoDAN-Turbo2026.01 | 12 |
| Method | Links | |
|---|---|---|
| AGENTICRED2026.01 | 60 | |
| CoP2026.01 | 38 | |
| AdvReasoning2026.01 | 36 | |
| J22026.01 | 35 | |
| AutoDAN-Turbo2026.01 | 12 |