List your company
ResearchFormerly WizWand

Image Understanding on Image benchmarks Aggregate

64.82Overall ScoreMRoPE-I
41.88847.841553.79559.7485Oct 27, 2025
Updated 6mo ago

Evaluation results

MethodLinks
MRoPE-IBackbone=Qwen3-VL-8B-Instruct2025.1064.82
MHRoPEBackbone=Qwen3-VL-8B-Instruct2025.1064.63
MRoPE-I2025.1063.79
CircleRoPEBackbone=Qwen3-VL-8B-Instruct2025.1063.54
RoPEBackbone=Qwen3-VL-8B-Instruct2025.1063.35
MRoPEBackbone=Qwen3-VL-8B-Instruct2025.1062.95
MHRoPE2025.1062.92
Vanilla RoPE2025.1062.17
MRoPE2025.1061.9
VideoRoPEBackbone=Qwen3-VL-8B-Instruct2025.1060.26
CircleRoPE2025.1060.16
HoPEBackbone=Qwen3-VL-8B-Instruct2025.1059.96
HoPE2025.1057.14
VideoRoPE2025.1057.03
MRoPE-IModel=Qwen3-VL-4B-Instruct2025.1048.82
MHRoPEModel=Qwen3-VL-4B-Instruct2025.1048.22
CircleRoPEModel=Qwen3-VL-4B-Instruct2025.1047.36
RoPEModel=Qwen3-VL-4B-Instruct2025.1047
MRoPEModel=Qwen3-VL-4B-Instruct2025.1046.67
HoPEModel=Qwen3-VL-4B-Instruct2025.1043.2
VideoRoPEModel=Qwen3-VL-4B-Instruct2025.1042.77