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

Multimodal Reasoning on MMBench (English and Chinese Scores)

84.29MMBench Accuracy (en)

Qwen3VL-2B-SFT

Updated 1d ago

Evaluation Results

Method
Qwen3VL-2B-SFT 2026.05	84.29	81.52	-
ETC 2026.05	82.75	81.44	-
ETC 2026.05	82.72	80.49	-
VoCo 2026.05	82.56	79.63	-
ETC 2026.05	82.4	80.39	-
VoCo 2026.05	82.01	79.83	-
VoCo 2026.05	81.34	78.91	-
Full 2026.04	66.1	58.9	-
LLaVA-1.5-7B 2025.03	66	58.9	100
D2-Pruning 2026.04	65.7	57.6	-
PreSel 2025.03	64.8	56.5	97.9
Vanilla 2026.05	64.6	58.1	100
TypiClust 2025.03	64.3	57.1	96.8
SPpruner 2026.05	63.8	57.6	99.4
DART 2026.05	63.7	57.3	99
SPpruner 2026.05	63.4	57	98.5
COINCIDE 2026.04	63.1	54.5	-
ICONS 2026.04	63.1	55.8	-
PyramidDrop 2026.05	63.1	57.1	98.7
DART 2026.05	63.1	57.3	97.2
PACT 2026.05	63	56.4	98.1
VisionZip 2026.05	62.9	57.4	99
DOSE 2026.04	62.5	54.8	-
DivPrune 2026.05	62.3	56.3	98.8
PyramidDrop 2026.05	62.3	56.6	96.5
Random 2026.04	62.2	54.8	-
PACT 2026.05	62.2	54.3	96.9
VisionZip 2026.05	62	56.7	97.6
SPpruner 2026.05	62	54.6	95.2
Self-Sup 2026.04	61.4	53.8	-
DART 2026.05	61.4	53.9	93.4
DivPrune 2026.05	61.3	54.9	97.4
FastV 2026.05	61.2	57	90
Random 2025.03	61	53.5	95.7
Qwen-VL-Chat 2025.03	60.6	56.7	-
COINCIDE 2025.03	60.5	53.9	95.5
PACT 2026.05	60.1	52.1	94
DivPrune 2026.05	60	52.5	94.4
Shikra 2025.03	58.8	-	-
EL2N 2025.03	58.2	48.5	93
PyramidDrop 2026.05	57.9	49.5	88.6
IFD 2025.03	57.2	50.6	91.8
FastV 2026.05	56.1	56.4	82.9
CLIP-Score 2026.04	55.2	52	-
VisionZip 2026.05	55.1	45.9	90.4
Perplexity 2025.03	54.7	47.8	91.9
IDEFICS-80B 2025.03	54.5	38.1	-
EL2N 2026.04	53.2	47.4	-
SemDeDup 2026.04	52.2	48.5	-
Perplexity 2026.04	52	45.8	-
CLIP-Score 2025.03	51	48	90.3
Self-Filter 2026.04	48.8	45.3	-
Self-Filter 2025.03	48.1	45.4	88.8
FastV 2026.05	48	52.7	74
InstructBLIP-7B 2025.03	36	23.7	-
Qwen3VL-2B-SFT 2026.05	35.74	35.09	-
InstructionGPT-4 2025.03	31.4	-	-
LLaMA-VID 2026.05	28.83	25.92	-
LLaMA-VID 2026.05	27.95	27.95	-
QueCC 2026.05	25.8	25.29	-
QueCC 2026.05	25.69	24.95	-