Share your thoughts, 1 month free Claude Pro on usSee more

Vision-Language Understanding on MMBench

88.7Accuracy

Qwen3-VL-4B-Instruct

Updated 4d ago

Evaluation Results

Method	Links
Qwen3-VL-4B-Instruct 2026.02		88.7
Xiaomi-Robotics-0 2026.02		84.4
MolmoAct 2026.02		80.1
Upper Bound 2026.04		68.9
CLASP 2026.04		68
CLASP 2026.04		68
Vanilla 2026.02		67.4
Vanilla 2025.05		67.4
SparseVLM 2026.04		67.4
PIO-FVLM 2026.02		66.2
MoB 2025.05		65.8
SparseVLM 2026.04		65.8
CDPruner 2026.02		65.5
DART 2026.02		65.3
HoloV 2026.02		65.3
Vanilla 2025.05		64.7
CLASP 2026.04		64.4
MoB 2025.05		64.1
TwigVLM 2025.05		64
MoB 2025.05		63.8
DART 2025.05		63.6
MoB 2025.05		63.5
PyramidDrop 2025.05		63.3
VisionZip 2026.02		63.1
VisionZip 2025.05		63
TokenCarve 2025.05		63
HiRED 2025.05		62.8
SparseVLM 2025.05		62.5
FiCoCo-V 2025.05		62.3
MustDrop 2025.05		62.3
MoB 2025.05		62.1
FastV 2026.02		61.6
SparseVLM 2026.04		61.3
FastV 2025.05		61.2
CLASP 2026.04		60.7
SparseVLM 2026.02		60.6
ToMe 2025.05		60.5
LLaVA-PruMerge 2025.05		59.6
SparseVLM 2026.04		56.2
pi_0.5 2026.02		22.1
pi_0 2026.02		0
Xiaomi-Robotics-0 2026.02		0