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

Multimodal Reasoning on MMBench (Accuracy, Acceptance Rate, and Speedup)

92.65Accuracy (%)

DREAM-R

14.07834.476554.87575.2735May 27, 2026
Updated 6d ago

Evaluation Results

MethodLinks
2026.05
92.6563.521.86
2026.05
90.463.81.62
2026.05
89.4453.282.48
2026.05
88.4571.452.31
2026.05
88.160.21.48
2026.05
86.9-1.05
2026.05
84.335.61.32
2026.05
83.9-1.11
2026.05
83.657.12.31
2026.05
83.2-1.06
2026.05
82.7-1.04
2026.05
82.3471.78
2026.05
82.237.961.73
2026.05
82.1-1.09
2026.05
81.9536.91.71
2026.05
81.7366.82.28
2026.05
81.3239.011.65
2026.05
81.31-1.09
2026.05
80.95-1.08
2026.05
80.566.421.93
2026.05
80.3269.991.77
2026.05
80.1552.62.18
2026.05
79.8537.91.62
2026.05
79.635.451.62
2026.05
79.4249.152.37
2026.05
79.11-1.15
2026.05
79.137.341.45
2026.05
78.2535.81.43
2026.05
77.3547.81.78
2026.05
76.4148.921.58
2026.05
74.735.461.46
2026.05
74.5271.751.73
2026.05
73.5832.131.28
2026.05
73.2968.471.13
2026.05
71.229.31.24
2026.05
69.7439.181.22
2026.05
67.137.51.2
2026.05
62.4924.261.43
2026.05
45.215.10.8
2026.05
17.114.820.71