Privacy Sanitization on PRIVASIS-SANITIZATION Vanilla (test)
91.53Sanitization Success Rate (Attribute)o3
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| o32026.02 | 91.53 | 91.62 | 74.57 | 94.06 | 95.24 | 93.57 | 70.25 | |
| GPT-52026.02 | 90.16 | 91.8 | 73.9 | 93.42 | 94.08 | 92.71 | 70.06 | |
| GPT-OSSParameters=120B2026.02 | 89.42 | 90.33 | 69.87 | 95.28 | 95.61 | 94.53 | 67.66 | |
| LLaMA-4 Maverick2026.02 | 89.31 | 90.77 | 73.61 | 89.97 | 90.27 | 88.68 | 67.18 | |
| DeepSeek R12026.02 | 88.54 | 91.01 | 72.26 | 93.74 | 94.39 | 93.38 | 69.58 | |
| GPT-4.12026.02 | 88.07 | 90.73 | 72.26 | 94.48 | 95.8 | 94.43 | 70.06 | |
| PRIVASIS-CLEANERParameters=4B2026.02 | 86.6 | 90.6 | 72.84 | 99.15 | 99.13 | 98.66 | 72.5 | |
| Qwen3Parameters=235B2026.02 | 83.34 | 87.58 | 68.04 | 93.47 | 94.31 | 93.19 | 64.4 | |
| PRIVASIS-CLEANERParameters=0.6B2026.02 | 81.61 | 87.37 | 68.52 | 99.26 | 99.23 | 98.75 | 68.04 | |
| Qwen3Parameters=4B2026.02 | 75.66 | 79.59 | 57.58 | 90.66 | 90.95 | 90.02 | 53.65 | |
| Qwen3Parameters=0.6B2026.02 | 47.39 | 53.77 | 35.99 | 70.49 | 71.42 | 69.19 | 16.7 |