GUI Perception and Reasoning on MPR-GUI-Bench TH
82.8AUGemini-2.5-Pro
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Gemini-2.5-ProModel Category=Close-source LVLMs2025.11 | 82.8 | 71.3 | 81.6 | 83.4 | 0.816 | 83.7 | 72 | 80 | 79.2 | |
| Gemini-1.5-FlashModel Category=Close-source LVLMs2025.11 | 77.9 | 79.5 | 67.7 | 86.3 | 0.59 | 40.4 | 64 | 48 | 63.5 | |
| Qwen-2.5-VL-7B-InstructModel Category=Open-source LVLMs2025.11 | 76.8 | 82.5 | 77.5 | 88.8 | 0.735 | 65.3 | 76 | 72 | 75.7 | |
| Keye-VL-7BModel Category=Open-source LVLMs2025.11 | 75.1 | 77.3 | 66 | 84.4 | 0.683 | 44 | 36 | 36 | 57 | |
| MiniCPM-o 2.6Model Category=Open-source LVLMs2025.11 | 66.4 | 74 | 60.6 | 78.1 | 0.631 | 42.6 | 52 | 40 | 57.1 | |
| Intern2.5VL-8BModel Category=Open-source LVLMs2025.11 | 57.9 | 67.5 | 52.6 | 72.7 | 0.429 | 38.3 | 80 | 52 | 58.5 | |
| CogAgent-9BModel Category=GUI Agents2025.11 | 50 | 69.7 | 54.8 | 68.6 | 0.32 | 31.1 | 40 | 20 | 42.8 |