List your company
ResearchFormerly WizWand

Multimodal Question Answering on MMBench-CN (dev)

73AccuracyQwen
60.20863.52966.8570.171Aug 1, 2025
Updated 6mo ago

Evaluation results

MethodLinks
QwenBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=100%2025.0873100
HiPruneBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=33.3%2025.0871.498.7
HiPrune++Backbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=33.3%2025.0871.198.7
FastVBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=33.3%2025.0870.697.4
VisionZipBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=33.3%2025.0869.897.7
HiPrune++Backbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=22.2%2025.0869.497.1
FastVBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=22.2%2025.0869.295.9
HiPruneBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=22.2%2025.0869.197.1
VisionZipBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=22.2%2025.0867.496.2
HiPruneBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=11.1%2025.0865.493
HiPrune++Backbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=11.1%2025.0864.793
VisionZipBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=11.1%2025.0863.391.5
FastVBackbone=Qwen2.5-VL-3B-Instruct, Token Retain Ratio=11.1%2025.0860.786.4