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

Multimodal Task Accuracy on OSWorld

41.49Multimodal Task Accuracy

KIVI

10.955618.882826.8134.7372May 16, 2026
Updated 15d ago

Evaluation Results

MethodLinks
2026.05
41.49
2026.05
41.43
2026.05
40.59
2026.05
40.54
2026.05
39.2
2026.05
35.78
2026.05
34.61
2026.05
34.02
2026.05
33.92
2026.05
29.52
2026.05
13.11
2026.05
12.77
2026.05
12.26
2026.05
12.22
2026.05
12.13