Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Multimodal Understanding on MMMU (Accuracy, Token Count, EoT)
Loading...
55.9
Accuracy
SketchThinker-R1-3B
52.052
53.051
54.05
55.049
Jan 6, 2026
Accuracy
Token Count
End-of-Turn Signal
Updated 4d ago
Evaluation Results
Method
Method
Links
Accuracy
Token Count
End-of-Turn Signal
SketchThinker-R1-3B
Backbone=Qwen2.5-VL-3B...
2026.01
55.9
54.5
1.026
Vanilla-R1
Backbone=Qwen2.5-VL-3B...
2026.01
54.8
128.3
0.427
C3oT
Backbone=Qwen2.5-VL-3B...
2026.01
54.1
107.5
0.503
Chain-of-Draft
Backbone=Qwen2.5-VL-3B...
2026.01
53.8
72.1
0.746
L1
Backbone=Qwen2.5-VL-3B...
2026.01
53.7
102.6
0.523
ThinkPrune
Backbone=Qwen2.5-VL-3B...
2026.01
53.2
95.2
0.559
Constrained CoT
Backbone=Qwen2.5-VL-3B...
2026.01
52.7
76.2
0.692
VeriThinker
Backbone=Qwen2.5-VL-3B...
2026.01
52.2
95.8
0.545
Feedback
Search any
task
Search any
task