Policy Corruption Evaluation on Gemini-2-Flash
3.65ComplianceHPM
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| HPMVictim Model=Gemini-2-Flash2025.12 | 3.65 | 3.33 | 3.01 | 3.75 | 2.92 | 3.05 | 3.11 | |
| PAPVictim Model=Gemini-2-Flash2025.12 | 2.31 | 2.15 | 1.9 | 2.35 | 1.33 | 1.62 | 1.55 | |
| CoAVictim Model=Gemini-2-Flash2025.12 | 1.92 | 1.6 | 1.42 | 2.03 | 0.95 | 0.91 | 1.12 | |
| PAIRVictim Model=Gemini-2-Flash2025.12 | 1.3 | 1.12 | 0.91 | 1.15 | 0.72 | 0.5 | 0.63 | |
| AutoDANVictim Model=Gemini-2-Flash2025.12 | 1.11 | 0.92 | 0.73 | 0.95 | 0.51 | 0.33 | 0.4 |