List your company
ResearchFormerly WizWand

Visual Question Answering on VQAv2 (test-std)

82.3AccuracyCoCa
70.8673.8376.879.77May 27, 2022Jan 17, 2023Sep 10, 2023May 3, 2024Dec 24, 2024Aug 17, 2025Apr 10, 2026
Updated 5mo ago

Evaluation results

MethodLinks
CoCaVocabulary=Closed2022.0582.3
FlamingoVocabulary=Open, Parameters=80B2022.0582.1
GIT2Vocabulary=Open, Parameters=5.1B2022.0581.92
mPlugVocabulary=Closed2022.0581.26
FlorenceVocabulary=Closed2022.0580.36
SimVLMVocabulary=Closed2022.0580.34
OFAVocabulary=Closed2022.0580.02
GITVocabulary=Open, Parameters=0.7B2022.0578.81
BLIPVocabulary=Closed2022.0578.32
METERVocabulary=Closed2022.0577.64
UncompressedModel=BLIP, Pruning Mode=/, Pruning Ratio=/, GFLOPs=186.12026.0477.5
UncompressedModel=BLIP, Pruning Mode=/, Pruning Ratio=/, GFLOPs=186.12026.0477.5
BLIPEvaluation protocol=Full fine-tuning, # Tunable=337M2025.1277.48
Null-LoRAEvaluation protocol=Frozen backbone, # Tunable=9.5M2025.1277.48
Ours‡Params. (M)=13.6, Mem. (G)=14.92026.0476.91
Ours†Params. (M)=10.9, Mem. (G)=12.62026.0476.87
Fully-FTParams. (M)=236.8, Mem. (G)=82.02026.0476.86
UFOVocabulary=Closed2022.0576.76
CLIP-ViLVocabulary=Closed2022.0576.7
VinVLVocabulary=Closed2022.0576.6
CoMPModel=BLIP, Pruning Mode=C, Pruning Ratio=0.7, GFLOPs=59.72026.0476.5
CoMPModel=BLIP, Pruning Mode=C, Pruning Ratio=0.7, GFLOPs=59.72026.0476.5
MADTPModel=BLIP, Pruning Mode=T, Pruning Ratio=0.7, GFLOPs=61.62026.0476.2
MADTPModel=BLIP, Pruning Mode=T, Pruning Ratio=0.7, GFLOPs=61.62026.0476.2
DoRAEvaluation protocol=Frozen backbone, # Tunable=10.8M, rank (r)=322025.1276.17
ALBEFEvaluation protocol=Full fine-tuning, # Tunable=266M2025.1276.04
ALBEFVocabulary=Closed2022.0576.04
SHERLParams. (M)=13.0, Mem. (G)=14.02026.0475.82
UniAdapterEvaluation protocol=Frozen backbone, # Tunable=19.5M, rank (r)=5122025.1275.56
UniPTParams. (M)=10.3, Mem. (G)=11.62026.0475.53
LSTParams. (M)=13.4, Mem. (G)=25.62026.0475.44
UNIMOVocabulary=Closed2022.0575.27
LoRAEvaluation protocol=Frozen backbone, # Tunable=10.6M, rank (r)=322025.1275.2
VILLAVocabulary=Closed2022.0574.87
UPopModel=BLIP, Pruning Mode=P, Pruning Ratio=0.7, GFLOPs=62.32026.0474.6
UPopModel=BLIP, Pruning Mode=P, Pruning Ratio=0.7, GFLOPs=62.32026.0474.6
PixelBERTVocabulary=Closed2022.0574.55
Visual ParsingVocabulary=Closed2022.0574.17
UNITEREvaluation protocol=Full fine-tuning, # Tunable=330M2025.1274.03
UNITERVocabulary=Closed2022.0574.02
OSCAREvaluation protocol=Full fine-tuning, # Tunable=330M2025.1273.82
OSCARVocabulary=Closed2022.0573.82
SOHOEvaluation protocol=Full fine-tuning, # Tunable=155M2025.1273.47
VL-T5/BARTEvaluation protocol=Full fine-tuning, # Tunable=165M2025.1271.3