Safety Evaluation on HVD Explicit
0HC RateClaude-3.7-Sonnet
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude-3.7-SonnetAttack=None2026.06 | 0 | 53.56 | |
| GPT-4oAttack=None2026.06 | 0.88 | 71.92 | |
| DeepSeek-V3Attack=None2026.06 | 3.54 | 82.85 | |
| GHOSTWRITERTarget Model=Claude-3.7-Sonnet, Attacker LM=GPT-4o-mini2026.06 | 62.83 | 92.73 | |
| GHOSTWRITERTarget Model=GPT-4o, Attacker LM=GPT-4o-mini2026.06 | 72.57 | 84.91 | |
| GHOSTWRITERTarget Model=DeepSeek-V3, Attacker LM=GPT-4o-mini2026.06 | 83.19 | 93 | |
| GHOSTWRITERTarget Model=Claude-3.7-Sonnet, Attacker LM=Qwen-2.5-7B2026.06 | 164.6 | 90.41 | |
| GHOSTWRITERTarget Model=GPT-4o, Attacker LM=Qwen-2.5-7B2026.06 | 202.65 | 87.12 | |
| GHOSTWRITERTarget Model=DeepSeek-V3, Attacker LM=Qwen-2.5-7B2026.06 | 258.41 | 93.42 |