AI Paper Atlas

Research map / Benchmarks / ImageNet-1k linear probe

ImageNet-1k linear probe leaderboard

14 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on ImageNet-1k linear probe. Best accuracy: 89.6 (ConDense, ECCV 2024).

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.

ImageNet-1k linear probe · accuracy

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

#accuracy ↑MethodPaperTopic
189.6ConDense
ViT-g/14
ConDense: Consistent 2D-3D Pre-training for Dense and Sparse Features from Multi-View Images
ECCV 2024 · Table 2
3D generation
286.5Web-DINO ViT-7B (224px)Scaling Language-Free Visual Representation Learning
ICCV 2025 · Table 3
Multimodal LLMs
386.29DINOv2-SINDER
DINOv2 giant
SINDER: Repairing the Singular Defects of DINOv2
ECCV 2024 · Table 3
Pruning and efficient tuning
485.2DINOv2 + attn. bias + ours
DINOv2 ViT-L
Revisiting [CLS] and Patch Token Interaction in Vision Transformers
ICLR 2026 · Table 2
Attention and transformers
583.75ALOE
DINOv3 ViT-B/16
Align Once to Explain: Feature Alignment for Scalable B-cosification of Foundational Vision Transformers
CVPR 2026 · Table 2
Attention and transformers
683.2Raptor k=4
DINOv2 ViT-B teacher
Block Recurrent Dynamics in Vision Transformers
ICLR 2026 · Table 1
Attention and transformers
782.4DSeq-JEPA
ViT-H/16 (448px, 300 epochs)
DSeq-JEPA: Discriminative Sequential Joint-Embedding Predictive Architecture
ECCV 2026 · Table 1
Self-supervised and contrastive
882.4Selective Aggregation (patches)
CAPI ViT-L
Beyond [cls]: Exploring the True Potential of Masked Image Modeling Representations
ICCV 2025 · Table 1
Self-supervised and contrastive
980.3Ours (PGT_F-B-1024)
PGT-B
Perceptual Group Tokenizer: Building Perception with Iterative Grouping
ICLR 2024 · Table 1
Self-supervised and contrastive
1079+StoP
ViT-H/14
Stochastic positional embeddings improve masked image modeling
ICML 2024 · Table: common MIM methods
Self-supervised and contrastive
4 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

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