Risk Detection on MobileRisk
86.92AccuracyAutoGLM
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| AutoGLMSafety Layer=VLM-as-critic2026.04 | 86.92 | 14.29 | 28.57 | 19.05 | 9.76 | |
| Gemini-3Type=Base2026.04 | 78.69 | 10.53 | 18.18 | 13.33 | 15.32 | |
| AutoGLMSafety Layer=Heuristic gate2026.04 | 76.47 | 11.54 | 15 | 13.04 | 15.33 | |
| AutoGLMSafety Layer=Prompt2026.04 | 74.86 | 13.33 | 17.39 | 15.09 | 16.67 | |
| GPT-5Type=Base2026.04 | 68.92 | 9.52 | 33.33 | 14.81 | 27.94 | |
| CORA2026.04 | 61.27 | 58.65 | 76.47 | 66.38 | 53.92 | |
| AutoGLM-VLMType=Base2026.04 | 58.82 | 62.5 | 44.12 | 51.72 | 26.47 | |
| UI-TARS-1.5Type=Base2026.04 | 50 | 50 | 6.86 | 12.07 | 6.86 |