List your company
ResearchFormerly WizWand

Multimodal Evaluation on MME (score)

2,375.9MME ScoreQwen-2.5-VL-7B + EvoLMM
1,685.4441,864.6972,043.952,223.203Aug 1, 2025Sep 20, 2025Nov 10, 2025Dec 31, 2025Feb 19, 2026Apr 11, 2026Jun 1, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Qwen-2.5-VL-7B + EvoLMMsetting=zero-shot2025.112,375.9
Qwen2.5-VL-7BToken Budget=100%2025.082,331
Qwen-2.5-VL-7Bsetting=zero-shot2025.112,318.8
HiPrune++Base Model=Qwen2.5-VL-7B, Token Budget=33.3%2025.082,310
HiPruneBase Model=Qwen2.5-VL-7B, Token Budget=33.3%2025.082,297
Upper BoundBackbone=InternVL2-26B2026.042,270
CLASPBackbone=InternVL2-26B, Token Retention=35%2026.042,262
ToMeBackbone=InternVL2-26B, Token Retention=35%2026.042,178
HiPrune++Base Model=Qwen2.5-VL-7B, Token Budget=22.2%2025.082,172
HiPruneBase Model=Qwen2.5-VL-7B, Token Budget=22.2%2025.082,168
Qwen2.5-VL-3BToken Budget=100%2025.082,144
FastVBackbone=InternVL2-26B, Token Retention=35%2026.042,140
HiPrune++Base Model=Qwen2.5-VL-3B, Token Budget=33.3%2025.082,069
HiPruneBase Model=Qwen2.5-VL-3B, Token Budget=33.3%2025.082,061
HiPruneBase Model=Qwen2.5-VL-3B, Token Budget=22.2%2025.082,002
HiPrune++Base Model=Qwen2.5-VL-7B, Token Budget=11.1%2025.081,999
HiPruneBase Model=Qwen2.5-VL-7B, Token Budget=11.1%2025.081,998
HiPrune++Base Model=Qwen2.5-VL-3B, Token Budget=22.2%2025.081,980
Upper BoundBase Model=LLaVA-NeXT-13B2026.041,901
LLaVA-NeXT-13BToken Num=28802025.081,901
HiPrune++Backbone=LLaVA-NeXT-13B, Token Num=6402025.081,894
HiPruneBase Model=Qwen2.5-VL-3B, Token Budget=11.1%2025.081,881
HiPruneBackbone=LLaVA-NeXT-13B, Token Num=6402025.081,877
CLASPBase Model=LLaVA-NeXT-13B, Token Retention Budget=640 Tokens2026.041,867
UncompressedPruning Mode=/, Pruning Ratio=/, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=5.632026.041,862
LLaVA-1.5-7BRetained Tokens=576, Pruning Ratio=100%2026.041,862
Upper BoundRetained Tokens=5762026.041,862
LLaVA-1.5-7BVTR Type=N/A, Tokens Retained=576, Reduction Ratio=100%2026.061,862
DARTPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=3.182026.041,856
Upper BoundBackbone=LLaVA-NeXT 7B, Tokens Retained=28802026.041,851
DeSAPRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,848
CLASPBase Model=LLaVA-NeXT-13B, Token Retention Budget=320 Tokens2026.041,848
CLASPRetained Tokens=192, Pruning Ratio=66.7%2026.041,848
PDropPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=3.092026.041,846
HiPrune++Base Model=Qwen2.5-VL-3B, Token Budget=11.1%2025.081,844
CoMPPruning Mode=C, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=2.942026.041,843
FlowCutRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,836
VisionZipPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=3.012026.041,834
DARTPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=3.182026.041,829
V2DropRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,826
DARTPruning Mode=T, Pruning Ratio=0.62, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=2.382026.041,823
HiPrune++Backbone=LLaVA-NeXT-13B, Token Num=3202025.081,823
SparseVLMBase Model=LLaVA-NeXT-13B, Token Retention Budget=640 Tokens2026.041,821
FastVVTR Type=Text-aware, Tokens Retained=192, Reduction Ratio=33.3%2026.061,821
HoloVRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,820
LLaVA-1.5-13BToken Num=5762025.081,818
VisPruner + RESTOREVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,816
DeSAPRetained Tokens=128, Pruning Ratio=↓ 77.8%2026.041,815
DivPrune + RESTOREVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,813
VisPruner + RESTOREVTR Type=Text-agnostic, Tokens Retained=128, Reduction Ratio=22.2%2026.061,813
HiPrune++Backbone=LLaVA-1.5-13B, Token Num=1922025.081,808
HoloV + RESTOREVTR Type=Text-agnostic, Tokens Retained=128, Reduction Ratio=22.2%2026.061,807
HoloVRetained Tokens=128, Pruning Ratio=↓ 77.8%2026.041,802
CLASPBase Model=LLaVA-NeXT-13B, Token Retention Budget=160 Tokens2026.041,801
HiPruneBackbone=LLaVA-NeXT-13B, Token Num=3202025.081,800
SparseVLMBase Model=LLaVA-NeXT-13B, Token Retention Budget=320 Tokens2026.041,798
HiPruneBackbone=LLaVA-1.5-13B, Token Num=1922025.081,798
PDropPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=3.092026.041,797
PyramidDropRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,797
DivPrune + RESTOREVTR Type=Text-agnostic, Tokens Retained=128, Reduction Ratio=22.2%2026.061,795
HoloV + RESTOREVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,793
FlowCutRetained Tokens=128, Pruning Ratio=↓ 77.8%2026.041,792
CLASPRetained Tokens=128, Pruning Ratio=77.8%2026.041,790
PDropVTR Type=Text-aware, Tokens Retained=192, Reduction Ratio=33.3%2026.061,790
SparseVLMPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=3.012026.041,787
HoloV + RESTOREVTR Type=Text-agnostic, Tokens Retained=64, Reduction Ratio=11.1%2026.061,787
VisPrunerVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,784
VisionZipPruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=3.012026.041,783
VisionZipRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,783
SparseVLMVTR Type=Text-aware, Tokens Retained=192, Reduction Ratio=33.3%2026.061,782
VisionZip + RESTOREVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,782
HoloVVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,779
VisPrunerVTR Type=Text-agnostic, Tokens Retained=128, Reduction Ratio=22.2%2026.061,778
HiPrune++Backbone=LLaVA-NeXT-13B, Token Num=1602025.081,774
VisionZip + RESTOREVTR Type=Text-agnostic, Tokens Retained=128, Reduction Ratio=22.2%2026.061,772
DARTPruning Mode=T, Pruning Ratio=0.62, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=2.382026.041,765
VisionZipRetained Tokens=128, Pruning Ratio=↓ 77.8%2026.041,763
DivPrunePruning Mode=T, Pruning Ratio=0.46, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=false, TFLOPs=3.012026.041,762
PyramidDropRetained Tokens=128, Pruning Ratio=↓ 77.8%2026.041,761
SparseVLMVTR Type=Text-aware, Tokens Retained=128, Reduction Ratio=22.2%2026.061,761
HoloVVTR Type=Text-agnostic, Tokens Retained=128, Reduction Ratio=22.2%2026.061,761
VisionZipPruning Mode=T, Pruning Ratio=0.62, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=2.162026.041,756
SparseVLMBase Model=LLaVA-NeXT-13B, Token Retention Budget=160 Tokens2026.041,755
VisionZipVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,749
DeSAPRetained Tokens=64, Pruning Ratio=↓ 88.9%2026.041,748
DivPrune + RESTOREVTR Type=Text-agnostic, Tokens Retained=64, Reduction Ratio=11.1%2026.061,748
HiPrune++Backbone=LLaVA-1.5-13B, Token Num=1282025.081,745
FlowCutRetained Tokens=64, Pruning Ratio=↓ 88.9%2026.041,744
HiPrune++Backbone=LLaVA-1.5-13B, Token Num=642025.081,736
Baseline: all KVVLM Backbone=Qwen3-VL-8B-Instruct2026.051,735.6
CoMPPruning Mode=C, Pruning Ratio=0.62, Backbone=LLaVA-v1.5-7B, Supervised fine-tuning=true, TFLOPs=2.072026.041,735
HiPruneBackbone=LLaVA-1.5-13B, Token Num=1282025.081,730
ToMeVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,727
VisionZip + RESTOREVTR Type=Text-agnostic, Tokens Retained=64, Reduction Ratio=11.1%2026.061,726
CLASPBackbone=LLaVA-NeXT 7B, Tokens Retained=320, Pruning Ratio=88.9%2026.041,723
DivPruneVTR Type=Text-agnostic, Tokens Retained=192, Reduction Ratio=33.3%2026.061,723
VisPruner + RESTOREVTR Type=Text-agnostic, Tokens Retained=64, Reduction Ratio=11.1%2026.061,722
SparseVLMRetained Tokens=192, Pruning Ratio=↓ 66.7%2026.041,721
HoloVRetained Tokens=64, Pruning Ratio=↓ 88.9%2026.041,715
V2DropRetained Tokens=128, Pruning Ratio=↓ 77.8%2026.041,712