Visual Question Answering On Vqa V2 Test Dev

Metrics

Accuracy

Results

Performance results of various models on this benchmark

Model Name	Accuracy	Paper Title	Repository
ONE-PEACE	82.6	ONE-PEACE: Exploring One General Representation Model Toward Unlimited Modalities
Pythia v0.3 + LoRRA	69.21	Towards VQA Models That Can Read
mPLUG (Huge)	82.43	mPLUG: Effective and Efficient Vision-Language Learning by Cross-modal Skip-connections
X-VLM (base)	78.22	Multi-Grained Vision Language Pre-Training: Aligning Texts with Visual Concepts
BEiT-3	84.19	Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks
Prismer	78.43	Prismer: A Vision-Language Model with Multi-Task Experts
CFR	72.5	Coarse-to-Fine Reasoning for Visual Question Answering
MUTAN	67.42	MUTAN: Multimodal Tucker Fusion for Visual Question Answering
Flamingo 80B	56.3	Flamingo: a Visual Language Model for Few-Shot Learning
Image features from bottom-up attention (adaptive K, ensemble)	69.87	Tips and Tricks for Visual Question Answering: Learnings from the 2017 Challenge
MMU	81.26	Achieving Human Parity on Visual Question Answering	-
ALBEF (14M)	75.84	Align before Fuse: Vision and Language Representation Learning with Momentum Distillation
Oscar	73.82	Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks
SimVLM	80.03	SimVLM: Simple Visual Language Model Pretraining with Weak Supervision
BLIP-2 ViT-G OPT 2.7B (zero-shot)	52.3	BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models
VK-OOD	77.9	Implicit Differentiable Outlier Detection Enable Robust Deep Multimodal Analysis
ViLT-B/32	71.26	ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision
MCAN+VC	71.21	Visual Commonsense R-CNN
BLIP-2 ViT-L FlanT5 XL (zero-shot)	62.3	BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models
BLIP-2 ViT-L OPT 2.7B (zero-shot)	49.7	BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models

0 of 56 row(s) selected.