Privacy Risk Evaluation on TOP-Bench (test)
80.95RLRQwen3-235B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen3-235Bmode=Instruct2025.12 | 80.95 | 37.62 | 0.292 | 47.78 | 32.51 | |
| MiniMax-M22025.12 | 87.14 | 28.57 | 0.218 | 61.17 | 25.73 | |
| Qwen3-30Bmode=Thinking2025.12 | 89.52 | 30.48 | 0.182 | 62.07 | 27.59 | |
| Qwen3-30Bmode=Instruct2025.12 | 90.95 | 41.43 | 0.157 | 51.72 | 38.92 | |
| DeepSeek-V3.2-Exp2025.12 | 92.38 | 32.86 | 0.137 | 60.68 | 31.55 | |
| Kimi-K2-Instruct2025.12 | 92.86 | 30.95 | 0.13 | 63.11 | 29.61 | |
| GLM-4.62025.12 | 93.33 | 31.43 | 0.122 | 62.62 | 30.58 | |
| Qwen3-235Bmode=Thinking2025.12 | 94.76 | 29.52 | 0.098 | 65.05 | 29.61 |