GUI Grounding on ScreenSpot v1 (test)
98.6Mobile Text AccGUI-G1
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GUI-G1Method Category=RL, Training Data Size=17K, Model Size=3B2026.02 | 98.6 | 85.8 | 96.4 | 80.7 | 91.4 | 82.3 | 90.3 | |
| Trifuse+GUI-AIMAMethod Category=Attention-based TB, Model Size=3B2026.02 | 98.2 | 86 | 96.9 | 85 | 94.8 | 79.1 | 90.6 | |
| Trifuse+GUI-ActorMethod Category=Attention-based TB, Model Size=7B2026.02 | 97.8 | 84.3 | 90.7 | 82.9 | 95.7 | 86.9 | 90.5 | |
| JEDIMethod Category=SFT, Training Data Size=1.39M, Model Size=3B2026.02 | 96.9 | 81.5 | 96.9 | 78.6 | 88.5 | 83.7 | 88.6 | |
| JEDIMethod Category=SFT, Training Data Size=1.39M, Model Size=7B2026.02 | 96.9 | 87.2 | 95.9 | 87.9 | 94.4 | 84.2 | 91.7 | |
| GUI-R1Method Category=RL, Training Data Size=3K, Model Size=3B2026.02 | 96.7 | 76.7 | 89.6 | 72.1 | 93.8 | 64.8 | 83.6 | |
| GUI-G2Method Category=RL, Training Data Size=100K, Model Size=7B2026.02 | 96.7 | 90.8 | 95.9 | 88.6 | 90.9 | 86.9 | 92 | |
| GUI-AIMAMethod Category=Attention-based TB, Training Data Size=85K, Model Size=3B2026.02 | 96.3 | 83.8 | 94.3 | 85.7 | 92.6 | 72.3 | 88.1 | |
| Trifuse+GUI-ActorMethod Category=Attention-based TB, Model Size=3B2026.02 | 96 | 82.5 | 90.7 | 80.7 | 96.5 | 85.4 | 89.5 | |
| UI-R1Method Category=RL, Training Data Size=136, Model Size=3B2026.02 | 95.6 | 84.7 | 85.2 | 73.3 | 90.2 | 59.3 | 83.3 | |
| GUI-ActorMethod Category=Attention-based TB, Training Data Size=1M, Model Size=7B2026.02 | 94.9 | 82.1 | 91.8 | 80 | 91.3 | 85.4 | 88.3 | |
| UI-TARSMethod Category=SFT, Training Data Size=50B Token, Model Size=7B2026.02 | 94.5 | 85.2 | 95.9 | 85.7 | 90 | 83.5 | 89.5 | |
| OmniParserMethod Category=General2026.02 | 93.9 | 57 | 91.3 | 63.6 | 81.3 | 51 | 73 | |
| Qwen2.5-VLMethod Category=General, Model Size=7B2026.02 | 93.4 | 76.4 | 87.6 | 57.9 | 82.2 | 63.1 | 78.6 | |
| UI-TARSMethod Category=SFT, Training Data Size=50B Token, Model Size=2B2026.02 | 93 | 75.5 | 90.7 | 68.6 | 84.3 | 74.8 | 82.3 | |
| GUI-ActorMethod Category=Attention-based TB, Training Data Size=1M, Model Size=3B2026.02 | 93 | 79.9 | 88.1 | 78.6 | 90.9 | 84 | 86.5 | |
| TrifuseMethod Category=Attention-based TF, Training Data Size=0, Model Size=7B2026.02 | 92.7 | 79 | 90.2 | 80 | 91.7 | 79.6 | 86.2 | |
| TrifuseMethod Category=Attention-based TF, Training Data Size=0, Model Size=3B2026.02 | 91.9 | 73.4 | 90.2 | 65.7 | 87.4 | 70.4 | 81.1 | |
| ShowUIMethod Category=SFT, Training Data Size=0.26M, Model Size=2B2026.02 | 91.6 | 69 | 81.8 | 59 | 83 | 65.5 | 74.9 | |
| TAGMethod Category=Attention-based TF, Training Data Size=0, Model Size=8.5B2026.02 | 88.3 | 29.3 | 82.5 | 28.6 | 70.9 | 29.1 | 57.5 | |
| MP-GUIMethod Category=SFT, Training Data Size=0.68M, Model Size=8B2026.02 | 86.8 | 65.9 | 70.8 | 56.4 | 58.3 | 46.6 | 64.1 | |
| UGroundMethod Category=SFT, Training Data Size=1.3M, Model Size=7B2026.02 | 82.8 | 60.3 | 82.5 | 63.6 | 80.4 | 70.4 | 73.3 | |
| AguviusMethod Category=SFT, Training Data Size=1M, Model Size=7B2026.02 | 78.2 | 88.3 | 70.7 | 88.1 | 74.8 | 85.7 | 81.8 | |
| Qwen2.5-VLMethod Category=General, Model Size=3B2026.02 | 62.1 | 46.4 | 54.1 | 30 | 31.2 | 48.3 | 46.9 | |
| GPT-4oMethod Category=General2026.02 | 30.5 | 23.2 | 20.6 | 19.4 | 11.1 | 7.8 | 18.8 |