Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

GUI Localization on GUIZoom-Bench hard-mislead

54.1Localization Accuracy

Qwen3-VL-32B

-2.16412.44327.0541.657Dec 5, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.12
54.1
2025.12
45.9
2025.12
45.9
2025.12
45.9
2025.12
40.5
2025.12
35.1
2025.12
35.1
2025.12
32.4
2025.12
29.7
2025.12
24.3
2025.12
24.3
2025.12
21.6
2025.12
21.6
2025.12
18.9
2025.12
18.9
2025.12
18.9
2025.12
18.9
2025.12
16.2
2025.12
16.2
2025.12
13.5
2025.12
10.8
2025.12
10.8
2025.12
10.5
2025.12
8.1
2025.12
5.4
2025.12
5.4
2025.12
2.7
2025.12
0
2025.12
0