Single-agent system security evaluation on OSWORLD (test)
36.7ASRCrescendo
Evaluation Results
| Method | Links | |
|---|---|---|
| CrescendoModel=GPT-4O-1120, System Type=Single-Agent2025.12 | 36.7 | |
| CrescendoModel=DOUBAO-VISION-PRO, System Type=Single-Agent2025.12 | 46.7 | |
| CrescendoModel=QWEN-VL-MAX, System Type=Single-Agent2025.12 | 53.3 | |
| Pop-upModel=QWEN-VL-MAX, System Type=Single-Agent2025.12 | 60 | |
| Pop-upModel=GPT-4O-1120, System Type=Single-Agent2025.12 | 66.7 | |
| Pop-upModel=DOUBAO-VISION-PRO, System Type=Single-Agent2025.12 | 76.7 |