List your company
ResearchFormerly WizWand

Multimodal Reasoning on MME

2,304Perception ScoreQwen2.5-VL-7B
1,206.281,491.2651,776.252,061.235Mar 11, 2026Mar 16, 2026Mar 22, 2026Mar 27, 2026Apr 2, 2026Apr 7, 2026Apr 13, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Qwen2.5-VL-7BToken Pruning Rate=0%2026.042,304
DeSAPToken Pruning Rate=66.7%2026.042,183
DeSAPToken Pruning Rate=77.8%2026.042,110
HoloVToken Pruning Rate=66.7%2026.042,093
FastVToken Pruning Rate=66.7%2026.042,072
DeSAPToken Pruning Rate=88.9%2026.042,056
HoloVToken Pruning Rate=77.8%2026.042,043
FastVToken Pruning Rate=77.8%2026.042,036
HoloVToken Pruning Rate=88.9%2026.042,006
FastVToken Pruning Rate=88.9%2026.041,940
Qwen2.5-VL-7BFT-Data=-, Fine-tuned on identical training data=false2026.031,693.9
SpaceR-sft-7BFT-Data=151K, Fine-tuned on identical training data=false2026.031,688.1
ViLASR-7BFT-Data=73K, Fine-tuned on identical training data=false2026.031,634.4
InternVL3-2BFT-Data=-, Fine-tuned on identical training data=false2026.031,610.2
Qwen2.5-VL-3BFT-Data=-, Fine-tuned on identical training data=false2026.031,526.1
Cambrain-S-3BFT-Data=10M, Fine-tuned on identical training data=false2026.031,490.8
GeoSenseFT-Data=940K, Fine-tuned on identical training data=false2026.031,473.7
VG-LLMFT-Data=385k, Fine-tuned on identical training data=false2026.031,441.6
SpatialLadder-3BFT-Data=26K, Fine-tuned on identical training data=false2026.031,403.2
Qwen2.5-VL-3B*FT-Data=940k, Fine-tuned on identical training data=true2026.031,301.2
VG-LLM*FT-Data=940k, Fine-tuned on identical training data=true2026.031,248.5