AI Paper Atlas

Research map / Video / Action recognition

V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learning

Lorenzo Mur Labadia, Matthew Muckley, Amir Bar, Mido Assran, Koustuv Sinha, Michael Rabbat, Yann LeCun, Nicolas Ballas, Adrien Bardes

ECCV 2026 · Video › Action recognition · approach: Contrastive video representation learning · 160 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add weighted context loss on visible tokens (ablation)V-JEPA 2 (ViT-L) · ADE20KmIoU: 22.2 -> 33.9Improved
add weighted context loss on visible tokens (ablation)V-JEPA 2 (ViT-L) · NYUv2RMSE: 0.682 -> 0.473Improved
add weighted context loss on visible tokens (ablation)V-JEPA 2 (ViT-L) · SSv2top-1 accuracy: 72.8 -> 62.5Hurt
add weighted context loss on visible tokens (ablation)V-JEPA 2 (ViT-L) · ImageNet-1Ktop-1 accuracy: 82.2 -> 72.6Hurt
add deep self-supervision at intermediate encoder layers (ablation)context-loss model · SSv2top-1 accuracy: 62.5 -> 72.0Improved
add deep self-supervision at intermediate encoder layers (ablation)context-loss model · ADE20KmIoU: 33.9 -> 38.6Improved
scale pretraining data to VisionMix163M (ablation)V-JEPA 2.1 ViT-L · ImageNet-1K / ADE20Kaccuracy / mIoU: 80.8 -> 81.6 on ImageNet, 38.6 -> 40.8 on ADE20KImproved
add multi-modal tokenizer (2D conv for images) (ablation)V-JEPA 2.1 ViT-L · ADE20KmIoU: 40.8 to 41.4Improved
scale encoder from ViT-L to ViT-G (ablation)V-JEPA 2.1 · ADE20KmIoU: 41.4 -> 47.1Improved
add high-resolution cool-down phase (ablation)V-JEPA 2.1 ViT-G · NYUv2RMSE: 0.365 -> 0.307Improved
distill ViT-G into ViT-L (ablation)V-JEPA 2.1 ViT-L · SSv2top-1 accuracy: 74.2% to 76.5%Improved
distill ViT-G into ViT-L (ablation)V-JEPA 2.1 ViT-L · ADE20KmIoU: 42.0 mIoU to 46.7 mIoUImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy