List your company

Vision-Language Evaluation on MME

2,498MME ScoreVanilla
1,204.241,540.121,8762,211.88May 15, 2025Jul 13, 2025Sep 11, 2025Nov 9, 2025Jan 8, 2026Mar 8, 2026May 7, 2026

Evaluation results

MethodLinks
VanillaBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=100%2026.052,498——
VisionZipBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.052,475——
VisMMOEBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.052,470——
MixBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.052,375——
VisionZipBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=50%2026.052,309——
VisMMOEBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=50%2026.052,299——
VanillaBackbone=DeepSeek-VL2, Pruning Ratio=100%2026.052,243——
MixBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=50%2026.052,212——
VisionZipBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.052,195——
RandomBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.052,181——
VisMMOEBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.052,178——
MixBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.052,153——
VisionZipBackbone=DeepSeek-VL2, Pruning Ratio=50%2026.052,142——
RandomBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.052,115——
VisMMOEBackbone=DeepSeek-VL2, Pruning Ratio=50%2026.052,103——
MixBackbone=DeepSeek-VL2, Pruning Ratio=50%2026.052,090——
RandomBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=50%2026.052,053——
RandomBackbone=DeepSeek-VL2, Pruning Ratio=50%2026.052,014——
VanillaBase Model=LLaVA-1.5-7B2025.051,862——
MoBBase Model=LLaVA-1.5-7B, Pruning budget K=192, η-prior=true2025.051,860——
VanillaBase Model=LLaVA-Next-7B2025.051,851——
Upper BoundBackbone=LLaVA-v1.5-13B, Retained Tokens=Full, Pruning Ratio=0%2026.041,818——
MoBBase Model=LLaVA-1.5-7B, Pruning budget K=64, η-prior=true2025.051,806——
CLASPBackbone=LLaVA-v1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.041,789——
CLASPBackbone=LLaVA-v1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.041,785——
SparseVLMBackbone=LLaVA-v1.5-13B, Retained Tokens=128, Pruning Ratio=77.8%2026.041,774——
SparseVLMBackbone=LLaVA-v1.5-13B, Retained Tokens=192, Pruning Ratio=66.7%2026.041,768——
MoBBase Model=LLaVA-Next-7B, Pruning budget K=320, η-prior=true2025.051,760——
CLASPBackbone=LLaVA-v1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.041,675——
SparseVLMBackbone=LLaVA-v1.5-13B, Retained Tokens=64, Pruning Ratio=88.9%2026.041,641——
CLASPBackbone=LLaVA-v1.5-13B, Retained Tokens=32, Pruning Ratio=94.4%2026.041,592——
BaselineRetain Tokens=Full2026.021,463——
SparseVLMBackbone=LLaVA-v1.5-13B, Retained Tokens=32, Pruning Ratio=94.4%2026.041,451——
IDPrunerRetain Tokens=1282026.021,428——
IDPrunerRetain Tokens=642026.021,329——
IDPrunerRetain Tokens=322026.021,254——
BLIP-22024.05—1,293.8290
IDEFICS2024.05—1,177.3—
InstructBLIP2024.05—1,212.8291.8
LLaVA-1.5-7BSIMA=false2024.05—1,506.4355.7
LLaVA-1.5-7BSIMA=true2024.05—1,507.7379.3
mPLUG-Owl22024.05—1,450.2313.2
Qwen-VL-Chat2024.05—1,487.6360.7
VILA-7BSIMA=false2024.05—1,533316.4
VILA-7BSIMA=true2024.05—1,559.6326.8