| Escaping the Big Data Paradigm with Compact Transformers | ✓ Link | 99.76 | | | | CCT-14/7x2 | 2021-04-12 |
| Reduction of Class Activation Uncertainty with Background Information | ✓ Link | 99.75 | | | | VIT-L/16 (Background) | 2023-05-05 |
| CvT: Introducing Convolutions to Vision Transformers | ✓ Link | 99.72 | | | | CvT-W24 | 2021-03-29 |
| Bamboo: Building Mega-Scale Vision Dataset Continually with Human-Machine Synergy | ✓ Link | 99.7 | | | | Bamboo (ViT-B/16) | 2022-03-15 |
| DINOv2: Learning Robust Visual Features without Supervision | ✓ Link | 99.7 | | | | DINOv2 (ViT-g/14) | 2023-04-14 |
| An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale | ✓ Link | 99.68 | | | | | 2020-10-22 |
| Enhance Image Classification via Inter-Class Image Mixup with Diffusion Model | ✓ Link | 99.68 | | | | Diff-Mix + CutMix (ViT-B/16@384) (extra data) | 2024-03-28 |
| Sharpness-Aware Minimization for Efficiently Improving Generalization | ✓ Link | 99.65% | | | | EffNet-L2 (SAM) | 2020-10-03 |
| Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision | ✓ Link | 99.65% | | | | ALIGN | 2021-02-11 |
| Big Transfer (BiT): General Visual Representation Learning | ✓ Link | 99.63 | | | | BiT-L (ResNet) | 2019-12-24 |
| Stabilizing Feature Geometry in Noisy Pretrained Models for Robust Downstream Tasks | | 99.54 | | | | FGS (Swin-L, ImageNet-21K) | 2026-06-03 |
| ConvMLP: Hierarchical Convolutional MLPs for Vision | ✓ Link | 99.5 | | | | ConvMLP-S | 2021-09-09 |
| ConvMLP: Hierarchical Convolutional MLPs for Vision | ✓ Link | 99.5 | | | | ConvMLP-L | 2021-09-09 |
| Effect of Pre-Training Scale on Intra- and Inter-Domain Full and Few-Shot Transfer Learning for Natural and Medical X-Ray Chest Images | ✓ Link | 99.49 | | | | ResNet-152x4-AGC (ImageNet-21K) | 2021-05-31 |
| AdapterTune: Zero-Initialized Low-Rank Adapters for Frozen Vision Transformers | ✓ Link | 99.43 | | | | AdapterTune (ViT-B/16) | 2026-03-16 |
| Big Transfer (BiT): General Visual Representation Learning | ✓ Link | 99.30 | | | | BiT-M (ResNet) | 2019-12-24 |
| SpinalNet: Deep Neural Network with Gradual Input | ✓ Link | 99.30 | | | | Wide-ResNet-101 (Spinal FC) | 2020-07-07 |
| Dense Vision Transformer Compression with Few Samples | | 99.3 | | | | DC-ViT-B (extra data) | 2024-03-27 |
| Multi-label Cluster Discrimination for Visual Representation Learning | ✓ Link | 99.3 | | | | MLCD ViT-L/14 (linear probe) | 2024-07-24 |
| Sensitivity-Aware Efficient Fine-Tuning via Compact Dynamic-Rank Adaptation | | 99.3 | | | | CDRA-SPT | 2025-06-11 |
| TResNet: High Performance GPU-Dedicated Architecture | ✓ Link | 99.1% | | | | TResNet-L | 2020-03-30 |
| Grafit: Learning fine-grained image representations with coarse labels | | 99.1% | | | | Grafit (RegNet-8GF) | 2020-11-25 |
| Going deeper with Image Transformers | ✓ Link | 99.1 | | | | CaiT-M-36 U 224 | 2021-03-31 |
| Adaptive Parameter Selection for Tuning Vision-Language Models | | 99.06 | | | | CLIP-AST (ViT-L/14) | 2025-06-11 |
| Domain Adaptive Transfer Learning on Visual Attention Aware Data Augmentation for Fine-grained Visual Categorization | | 98.9% | | | | DAT | 2020-10-06 |
| Image Captioners Are Scalable Vision Learners Too | ✓ Link | 98.9 | | | | CapPa L/14 + MAP | 2023-06-13 |
| HAMSA: Scanning-Free Vision State Space Models via SpectralPulseNet | ✓ Link | 98.9 | | | | Hamsa-B | 2026-04-16 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 98.8% | | | | EfficientNet-B7 | 2019-05-28 |
| Training data-efficient image transformers & distillation through attention | ✓ Link | 98.8% | | 86M | | DeiT-B | 2020-12-23 |
| EfficientNetV2: Smaller Models and Faster Training | ✓ Link | 98.8 | | | | EfficientNetV2-L | 2021-04-01 |
| Global Filter Networks for Image Classification | ✓ Link | 98.8 | | 54M | | GFNet-H-B | 2021-07-01 |
| Incorporating Convolution Designs into Visual Transformers | ✓ Link | 98.6 | | | | CeiT-S (384 finetune resolution) | 2021-03-22 |
| EfficientNetV2: Smaller Models and Faster Training | ✓ Link | 98.5 | | | | EfficientNetV2-M | 2021-04-01 |
| Three things everyone should know about Vision Transformers | ✓ Link | 98.5 | | | | ViT-B (attn finetune) | 2022-03-18 |
| HAMSA: Scanning-Free Vision State Space Models via SpectralPulseNet | ✓ Link | 98.4 | | | | Hamsa-S | 2026-04-16 |
| Neural Architecture Transfer | ✓ Link | 98.3% | 400M | 4.2M | | NAT-M4 | 2020-05-12 |
| LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference | ✓ Link | 98.3 | | | | LeViT-384 | 2021-04-02 |
| Effect of Pre-Training Scale on Intra- and Inter-Domain Full and Few-Shot Transfer Learning for Natural and Medical X-Ray Chest Images | ✓ Link | 98.21 | | | | ResNet-50x1-ACG (ImageNet-21K) | 2021-05-31 |
| Incorporating Convolution Designs into Visual Transformers | ✓ Link | 98.2 | | | | CeiT-S | 2021-03-22 |
| Neural Architecture Transfer | ✓ Link | 98.1% | 250M | 3.7M | | NAT-M3 | 2020-05-12 |
| Neural Architecture Transfer | ✓ Link | 97.9% | 195M | 3.4M | | NAT-M2 | 2020-05-12 |
| EfficientNetV2: Smaller Models and Faster Training | ✓ Link | 97.9 | | | | EfficientNetV2-S | 2021-04-01 |
| ResMLP: Feedforward networks for image classification with data-efficient training | ✓ Link | 97.9 | | | | ResMLP24 | 2021-05-07 |
| ResNet strikes back: An improved training procedure in timm | ✓ Link | 97.9 | 4.1 | 25M | | ResNet50 (A1) | 2021-10-01 |
| TransBoost: Improving the Best ImageNet Performance using Deep Transduction | ✓ Link | 97.85% | | | | TransBoost-ResNet50 | 2022-05-26 |
| Incorporating Convolution Designs into Visual Transformers | ✓ Link | 97.8 | | | | CeiT-T (384 finetune resolution) | 2021-03-22 |
| LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference | ✓ Link | 97.8 | | | | LeViT-192 | 2021-04-02 |
| ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Bias | ✓ Link | 97.8 | | | | ViTAE-S | 2021-06-07 |
| LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference | ✓ Link | 97.7 | | | | LeViT-256 | 2021-04-02 |
| ResMLP: Feedforward networks for image classification with data-efficient training | ✓ Link | 97.4 | | | | ResMLP12 | 2021-05-07 |
| Classification-Specific Parts for Improving Fine-Grained Visual Categorization | ✓ Link | 96.9% | | | | CS-Parts | 2019-09-16 |
| Incorporating Convolution Designs into Visual Transformers | ✓ Link | 96.9 | | | | CeiT-T | 2021-03-22 |
| LeViT: a Vision Transformer in ConvNet's Clothing for Faster Inference | ✓ Link | 96.8 | | | | LeViT-128S | 2021-04-02 |
| FLAVA: A Foundational Language And Vision Alignment Model | ✓ Link | 96.37 | | | | FLAVA | 2021-12-08 |
| Vision Models Are More Robust And Fair When Pretrained On Uncurated Images Without Supervision | ✓ Link | 96.3 | | | | SEER (RegNet10B) | 2022-02-16 |
| MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning | ✓ Link | 96.2 | | | | MULTIFLOW (DINO_B/16, 63% sparsity) (extra data) | 2024-04-08 |
| With a Little Help from My Friends: Nearest-Neighbor Contrastive Learning of Visual Representations | ✓ Link | 95.1 | | | | NNCLR | 2021-04-29 |
| $λ$-JEPA Spectral Anti-Collapse Regularization for Self-Supervised Learning | ✓ Link | 92.2 | | | | λ-JEPA ViT-B/16, 100 ep (linear probe) | 2026-09-28 |
| Closest Neighbors are Harmful for Lightweight Masked Auto-encoders | ✓ Link | 91.85 | | | | NoR-MAE (ViT-Small) | 2025-06-11 |
| When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations | ✓ Link | 91.8 | | | | ViT-B/16- SAM | 2021-06-03 |
| When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations | ✓ Link | 91.5 | | | | ViT-S/16- SAM | 2021-06-03 |
| When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations | ✓ Link | 91.1 | | | | ResNet-152-SAM | 2021-06-03 |
| When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations | ✓ Link | 90 | | | | ResNet-50-SAM | 2021-06-03 |
| When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations | ✓ Link | 90 | | | | Mixer-B/16- SAM | 2021-06-03 |
| Linear Attention with Global Context: A Multipole Attention Mechanism for Vision and Physics | ✓ Link | 89.00 | | | | MANO-tiny | 2025-07-03 |
| When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations | ✓ Link | 87.9 | | | | Mixer-S/16- SAM | 2021-06-03 |
| Your Diffusion Model is Secretly a Zero-Shot Classifier | ✓ Link | | | | 66.3 | Diffusion Classifier (zero-shot) | 2023-03-28 |
| Neural Architecture Transfer | ✓ Link | | 152M | 3.3M | | NAT-M1 | 2020-05-12 |
| GeoStack: A Framework for Quasi-Abelian Knowledge Composition in VLMs | ✓ Link | 85.8 | | | | GeoStack | 2026-05-07 |
| FDSML: Formula-Driven Supervised Metric Learning with Parameter-Aware Triplet Loss | ✓ Link | 74.21 | | | | FDSML (ViT-Tiny, each) (extra data) | 2026-10-29 |
| HeBA: Heterogeneous Bottleneck Adapters for Robust Vision-Language Models | ✓ Link | 73.04 | | | | HeBA | 2026-03-17 |