AI Paper Atlas

Research map / Benchmarks / VQA

VQA leaderboard

28 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on VQA. Best accuracy (test-dev): 81.11 (mPLUG-2, ICML 2023).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

VQA · accuracy (test-dev)

23 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
181.11mPLUG-2mPLUG-2: A Modularized Multi-modal Foundation Model Across Text, Image and Video
ICML 2023 · Table: visual question answeri
Cross-modal learning
278.44PTP-BLIP (14M)
BLIP
Position-Guided Text Prompt for Vision-Language Pre-Training
CVPR 2023 · VQA and NLVR2 results table
CLIP and zero-shot
378.34BUS 512
ViT-B/16, 4M pretrain
BUS: Efficient and Effective Vision-Language Pre-Training with Bottom-Up Patch Summarization.
ICCV 2023 · Table 1
Human and embodied modeling
478.33OmniVL (14M image-text + video)OmniVL: One Foundation Model for Image-Language and Video-Language Tasks
NeurIPS 2022 · Table 8
Human and embodied modeling
578.22X-VLM (16M)
X-VLM (4M)
Multi-Grained Vision Language Pre-Training: Aligning Texts with Visual Concepts
ICML 2022 · Downstream V+L tasks table
Human and embodied modeling
676.52Oscar+ with VinVL (Large)
BERT-Large
VinVL: Revisiting Visual Representations in Vision-Language Models
CVPR 2021 · Table: overall comparison with
Open-set and few-shot detection
776.11Knowledge-CLIP
CLIP ViT-L/14
Contrastive Language-Image Pre-Training with Knowledge Graphs
NeurIPS 2022 · Table 2
CLIP and zero-shot
876.04Geo-MAFA
MAFA
GeoMM: On Geodesic Perspective for Multi-modal Learning
CVPR 2025 · Table 3
Graph algorithms and clustering
975.69NARVL base
OFA base
Non-autoregressive Sequence-to-Sequence Vision-Language Models
CVPR 2024 · Table 1
Human and embodied modeling
1075.55MAFA
GRIT-VLP
MAFA: Managing False Negatives for Vision-Language Pre-training
CVPR 2024 · Table 3
CLIP and zero-shot
13 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

VQA · accuracy (test-std)

10 papers, each with its best reported number for its own method; higher is better.

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.