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

Multi-image spatial reasoning on SPBench-MV

80.4Accuracy

GAMSI_S1+S2

40.25650.67861.171.522May 25, 2026
Updated 8d ago

Evaluation Results

MethodLinks
2026.05
80.4
2026.05
76.9
2026.05
71.7
2026.05
71.1
2026.05
70.3
2026.05
68.5
65.7
2026.05
65
2026.05
57.5
2026.05
56.7
2026.05
56.2
55.9
2026.05
55.4
2026.05
54.4
52.9
46.7
43.1
42.7
41.8