List your company
ResearchFormerly WizWand

Visual Question Answering on GQA (val)

83.39AccuracyOracle
47.832457.063766.29575.5263May 10, 2019Jun 30, 2020Aug 21, 2021Oct 12, 2022Dec 3, 2023Jan 23, 2025Mar 17, 2026
Updated 7mo ago

Evaluation results

MethodLinks
Oracle2026.0383.39
PEVLgrounded inputs=true2022.0577
CFR2021.1073.6
CFR2022.0573.6
LCStraining_mode=learned, evaluation_protocol=5-fold cross-validation2026.0366.47
LLaVA-NeXT-INST-ITLLM=Vicuna-7B, Vision Encoder=CLIP-ViT-Large2024.1265.9
LLaVA-NeXT-INST-ITLLM=Qwen2-7B, Vision Encoder=SigLIP-SO4002024.1265.5
DeepStack-L-HDLLM=Vicuna-7B, Vision Encoder=CLIP-ViT-Large2024.1265.2
FAAR-learntraining_mode=learned, evaluation_protocol=5-fold cross-validation2026.0364.89
ALBEF2022.0564.8
QualRCCVrho=0.4, gamma=1, training_mode=training-free2026.0364.33
RCCVrho=0.4, training_mode=training-free2026.0364.3
HFV-sharptraining_mode=training-free2026.0364.27
Single bestpool_size=17 models, families=8 families2026.0364.25
LLaVA-NeXTLLM=Vicuna-7B, Vision Encoder=CLIP-ViT-Large2024.1264.2
HFVtraining_mode=training-free2026.0364.18
Calibrated votepool_size=17 models, families=8 families2026.0364.02
single-hop + LCGNVisual features=Object detection features, Message passing rounds (T)=42019.0563.9
Majority votepool_size=17 models, families=8 families2026.0363.72
DeepStack-LLLM=Vicuna-7B, Vision Encoder=CLIP-ViT-Large2024.1263.1
VILALLM=Vicuna-7B, Vision Encoder=CLIP-ViT-Large2024.1262.3
single-hopVisual features=Object detection features2019.0562
LLaVA-1.5LLM=Vicuna-7B, Vision Encoder=CLIP-ViT-Large2024.1262
CTI2021.1061.7
CTI2022.0561.7
BAN2021.1061.5
BAN2022.0561.5
LXMERT2021.1059.8
LXMERT2022.0559.8
MACVisual features=Object detection features2019.0557.5
Bottom-UpVisual features=Object detection features2019.0552.2
CNN+LSTMVisual features=Convolutional grid features2019.0549.2