List your company
ResearchFormerly WizWand

Natural Language Visual Reasoning on NLVR2 (test-p)

92.6AccuracyBEIT-3
81.742484.561287.3890.1988Jan 2, 2021Sep 24, 2021Jun 17, 2022Mar 10, 2023Dec 1, 2023Aug 23, 2024May 16, 2025
Updated 7mo ago

Evaluation results

MethodLinks
BEIT-32023.0592.6
BEIT-32022.0892.58
VLMO-Large++# Pretrain Images=1.0B2021.1189.54
X2-VLM_large# Params=593M, Pre-training Data=More Data2022.1189.4
ONE-PEACE2023.0588.3
X2-VLM_large# Params=593M, Pre-training Data=4M2022.1187.6
CoCaInput Modality=image only, Evaluation Protocol=lightweight finetuning2022.0587
CoCa2022.0887
CoCa2023.0587
CoCa#Images=3B, Para.=2.1B2022.1287
X2-VLM_base# Params=255M, Pre-training Data=More Data2022.1187
VLMoInput Modality=image only, Evaluation Protocol=lightweight finetuning2022.0586.9
VLMo_large# Params=562M, Pre-training Data=4M2022.1186.9
VLMO-Large# Pretrain Images=4M, Model Size=Large2021.1186.86
VLMo-LPretrain Images=4M2022.0686.86
VLMo2022.0586.86
X2-VLM_base# Params=255M, Pre-training Data=4M2022.1186.1
FIBER-BPretrain Images=4M2022.0685.52
SimVLMInput Modality=image only, Evaluation Protocol=lightweight finetuning2022.0585.2
SimVLM2023.0585.2
SimVLMModel Size=Huge2021.0885.15
SimVLM-Huge# Pretrain Images=1.8B2021.1185.15
SimVLM2022.0885.15
SimVLM-HPretrain Images=1.8B2022.0685.15
mPLUG2022.0584.95
SimVLMModel Size=Large2021.0884.84
SimVLM-Large# Pretrain Images=1.8B2021.1184.84
SimVLM-LARGE# Pre-train Images=1.8B, Model scale=Large2023.0584.84
SimVLMSize=Large2022.0584.84
SimVLM_large# Params=783M, Pre-training Data=More Data2022.1184.8
X-VLMPre-training Images Scale=16M2021.1184.76
X-VLMPretrain Images=4M2022.0684.76
X-VLM2023.0584.76
SCLPre-training image scale=<10M images2022.1184.27
X-VLMPre-training Images Scale=4M2021.1184.21
SimVLM-BASE# Pre-train Images=1.8B, Model scale=Base2023.0584.15
VinVLInput Modality=image only, Evaluation Protocol=lightweight finetuning2022.0584
OSCAR+ w/ VinVLModel Scale=large2021.0183.98
VinVLModel Size=Large2021.0883.98
VinVL-Large# Pretrain Images=5.7M2021.1183.98
VinVL_largePre-training Images Scale=5.6M2021.1183.98
VinVL2022.0883.98
VinVL_L#Images=5.6M, Para.=347M2022.1283.98
MAPPre-training dataset=<10M images, Model size=Base size2022.1083.48
METER-Swin-BPretrain Images=4M2022.0683.47
VLMO-Base# Pretrain Images=4M, Model Size=Base2021.1183.34
VLMo-BPretrain Images=4M2022.0683.34
VLMoPre-training image scale=<10M images2022.1183.34
VLMo-BASE# Pre-train Images=4M, Model scale=Base2023.0583.34
ManagerTower-BASE# Pre-train Images=4M, Model scale=Base2023.0583.34
VLMo-BasePre-training dataset=<10M images, Model size=Base size2022.1083.34
VLMo_base# Params=175M, Pre-training Data=4M2022.1183.3
CCLM_basePre-training Data=4M2022.0683.22
PTP-BLIP#Images=14M, Para.=220M2022.1283.17
ALBEFPre-training Images=14M2021.0783.14
ALBEFPre-train #Images=14M2022.0183.14
ALBEFPre-training Images Scale=14M2021.1183.14
ALBEF2022.0883.14
ALBEF(14M)Pre-training image scale=>10M images2022.1183.14
ALBEF-BASE (14M)# Pre-train Images=14M, Model scale=Base2023.0583.14
ALBEF2023.0583.14
ALBEF#Images=14M, Para.=210M2022.1283.14
ALBEF2022.0583.14
SimVLM-BasePre-training dataset=>10M images, Model size=Base size2022.1083.14
ALBEFPre-train # Images=14M2023.1283.14
ALBEFInput Modality=image only, Evaluation Protocol=lightweight finetuning2022.0583.1
VinVL#Img=6M, Encoder Type=Object Detector (O.D.)2023.0483.1
VinVLSupervision Level=Supervised, Pre-training Data=Large-Scale Paired Image-Text Data2023.0583.1
METER# Params=341M, Pre-training Data=4M2022.1183.1
BLIP_base# Params=240M, Pre-training Data=More Data2022.1183.1
BridgeTower-BASE# Pre-train Images=4M, Model scale=Base2023.0583.09
OSCAR+ w/ VinVLModel Scale=base2021.0183.08
VinVL-BaseVisual Embed=Region, Time (ms)=~650, Additional pre-training data=GQA, VQAv2, VG-QA, Open Images2021.0283.08
BLIPPre-train #Images=129M2022.0183.08
VinVL#Images=6M2022.0283.08
BLIP-BASE# Pre-train Images=129M, Model scale=Base2023.0583.08
VinVL_baseModel Scale=base2022.0683.08
VinVL-BasePre-training dataset=<10M images, Model size=Base size2022.1083.08
METER-CLIP2021.1183.05
METERPre-training image scale=<10M images2022.1183.05
METER-CLIP-BASE# Pre-train Images=4M, Model scale=Base2023.0583.05
METER2022.0583.05
METERPre-training dataset=<10M images, Model size=Base size2022.1083.05
METER-CLIP-VITBASESupervision Level=Supervised, Pre-training Data=Large-Scale Paired Image-Text Data2023.0583
VL-BEiT# Params=175M, Pre-training Data=4M2022.1182.7
CoCa-LargeParams (million)=672.12023.1082.6
METER-Swin2021.1182.47
METER-Swin-BASE# Pre-train Images=4M, Model scale=Base2023.0582.47
BLIPPre-train #Images=14M2022.0182.3
BLIP#Images=14M, Para.=220M2022.1282.3
BLIP2022.0582.3
BLIPPre-train # Images=14M2023.1282.3
FALCON#Images=4M2025.0582.28
BLIPCapFilt-LPre-train #Images=129M2022.0182.24
BLIP2022.0882.24
BLIP_CapFilt-LPretrain Images=129M2022.0682.24
BLIPPre-training image scale=>10M images2022.1182.24
BLIP2023.0582.24
BLIPInput Modality=image only, Evaluation Protocol=lightweight finetuning2022.0582.2
MAFAPre-train # Images=4M, Pre-train Dataset=4M-Clean2023.1282.16