AI Paper Atlas

Research map / Benchmarks / NLVR² test-P

NLVR² test-P leaderboard

10 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on NLVR² test-P. Best accuracy: 85.52 (FIBER-B, NeurIPS 2022).

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.

NLVR² test-P · accuracy

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

#accuracy ↑MethodPaperTopic
185.52FIBER-B
Swin-Base + RoBERTa-Base
Coarse-to-Fine Vision-Language Pre-training with Fusion in the Backbone
NeurIPS 2022 · Table: Results on VL classific
Human and embodied modeling
283.14ALBEF (14M)
ViT-B/16 + BERT_base
Align before Fuse: Vision and Language Representation Learning with Momentum Distillation
NeurIPS 2021 · Table: Comparison with state-o
Human and embodied modeling
382.3BLIPBLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation
ICML 2022 · Table: VQA and NLVR²
CLIP and zero-shot
481.6SoftMask++Multi-Modal Representation Learning With Text-Driven Soft Masks
CVPR 2023 · Table 3
Human and embodied modeling
581.33TCLVision-Language Pre-Training With Triple Contrastive Learning
CVPR 2022 · Table 4
Cross-modal learning
680.84CODIS
ViT-B/16, BERTbase
Multi-Modal Alignment Using Representation Codebook
CVPR 2022 · Table: downstream vision-langu
Cross-modal learning
780.6OursWebly Supervised Knowledge Embedding Model for Visual Reasoning
CVPR 2020 · Table 2
Human and embodied modeling
877.32SOHO
ResNet-101
Seeing Out of the Box: End-to-End Pre-Training for Vision-Language Representation Learning
CVPR 2021 · NLVR² table
Human and embodied modeling
974.2CLIP-BART + Single Adapter
CLIP-BART
VL-Adapter: Parameter-Efficient Transfer Learning for Vision-and-Language Tasks
CVPR 2022 · Table 1
CLIP and zero-shot
1073.6VL-T5
T5Base
Unifying Vision-and-Language Tasks via Text Generation
ICML 2021 · Table: Single model performanc
Visual question answering

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