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

Multi-image visual reasoning on BLINK

69.1Accuracy

Qwen3-VL

17.93231.21644.557.784Sep 30, 2025Nov 14, 2025Dec 29, 2025Feb 13, 2026Mar 30, 2026May 14, 2026Jun 29, 2026
Updated 25d ago

Evaluation Results

MethodLinks
2026.05
69.1
2026.06
64.5
2026.06
64.3
2026.06
64.2
2026.06
63.1
2026.05
62.8
2025.09
60.1
2026.05
59.5
2025.09
57
2026.05
56.7
2026.04
55.85
2025.09
55.3
2026.04
54.29
2026.04
53.92
2026.05
53.9
2026.05
53.8
2026.04
51.61
2026.05
51.3
2026.04
51.1
2026.06
46.93
2026.04
45.2
2026.05
44.2
2026.06
43.93
2026.04
43.4
2026.04
42.9
2026.04
42.2
2026.04
41.5
2026.04
41.28
2026.04
40.8
2026.06
40.6
2026.04
40.5
2026.06
39.8
2026.04
39.7
2026.06
39.7
2026.04
39.6
2026.06
39.6
2026.06
39.6
2026.04
38.9
2026.04
38.6
2026.06
38.6
2026.06
38.5
2026.04
38.3
2026.04
38.2
2026.04
37.1
2026.04
36.6
2026.06
36.39
2026.04
36.2
2026.04
31.2
2026.04
28.2
2026.04
27.9
2026.04
19.9