Multimodal Benchmarking on MMBench (Score and Relative Performance)

83.33MMBench Score

Training-Free Debiasing Inference Strategy

Updated 22d ago

Evaluation Results

Method
Training-Free Debiasing Inference Strategy 2026.05	83.33	-
DMAS 2026.05	83.2	-
VISTA 2026.05	82.7	-
VCD 2026.05	82.55	-
SID 2026.05	82.15	-
Zero-shot 2026.05	81.13	-
Full Model 2026.04	72.2	-
TopoVLM 2026.04	68.1	-
TAMP 2026.04	66.5	-
LLM-Streamline 2026.04	66.3	-
ECoFLaP 2026.04	66.2	-
SparseGPT 2026.04	65.7	-
PDrop+E 2026.03	65.2	101.4
PDrop 2026.03	65	101.1
FastV+E 2026.03	64.7	100.6
Vanilla 2026.04	64.7	100
Base 2026.03	64.3	-
VisionZip+E 2026.03	64.1	99.7
LLM-Pruner 2026.04	64.1	-
FastV 2026.03	64	99.5
VisionZip 2026.03	64	99.5
FastV 2026.03	63.8	99.2
Wanda 2026.04	63.8	-
LearnPruner 2026.04	63.8	98.5
Training-Free Debiasing Inference Strategy 2026.05	63.7	-
SID 2026.05	63.65	-
TwigVLM 2026.04	63.5	99
DMAS 2026.05	63.3	-
VCD 2026.05	63.1	-
FastV+E 2026.03	63	98
PDrop 2026.03	63	98
VisPruner 2026.04	62.7	97.3
VISTA 2026.05	62.7	-
VisionZip ‡ 2026.04	62.6	97.3
LearnPruner 2026.04	62.6	96.9
VisionZip+E 2026.03	62.5	97.2
VisionZip 2026.03	62.2	96.7
PDrop+E 2026.03	62	96.4
DivPrune 2026.04	62	96.4
VisionZip 2026.04	62	96.3
Zero-shot 2026.05	62	-
VisionZip ‡ 2026.04	61.5	95.1
VisPruner 2026.04	61.3	94.3
OWL 2026.04	61	-
LearnPruner 2026.04	60.8	94.8
DART 2026.04	60.7	95.4
TwigVLM 2026.04	60.4	96
VisionZip 2026.04	60.1	93
SparseVLM 2026.04	60	92.6
DART 2026.04	59.5	91.9
DivPrune 2026.04	59.3	93.9
DART 2026.04	58.5	88
VisPruner 2026.04	58.4	89.7
VisionZip 2026.04	57.7	87.8
DivPrune 2026.04	57.6	90.5
SparseVLM 2026.04	56.2	85.6
FastV 2026.04	56.1	82.1
SparseVLM 2026.04	51.4	77.5
FastV 2026.04	48	71.4
FastV 2026.04	37.8	58.6