AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Rethinking JEPA: Compute‑Efficient Video Self-Supervised Learning with Frozen Teachers

Xianhang Li, Chen Huang, Chun-Liang Li, Eran Malach, Joshua M. Susskind, Vimal Thilak, Etai Littwin

ICLR 2026 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning

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
replace EMA teacher with frozen pixel-reconstruction teacher (SALT) (ablation)V-JEPA 2 ViT-L on V-3.6M · 6-benchmark averageavg top-1 accuracy: 79.6 -> 81.9Improved
replace EMA teacher with frozen teacher (SALT)V-JEPA 2 ViT-L on V-3.6M · SSv2 (16x2x3)top-1 accuracy: 68.2 -> 74.9Improved
train student with stage 2 latent prediction after teacher (ablation)SALT ViT-L · 6-benchmark averageavg top-1 accuracy: 77.3 -> 81.9Improved
use ViT-L teacher instead of larger ViT-H/ViT-G teacher (ablation)SALT ViT-L student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 77.4 vs 77.3 and 77.6No clear effect
use ViT-L teacher instead of ViT-H teacher (ablation)SALT ViT-G student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 78.7 -> 79.0Improved
teacher pretrained on ImageNet-1k instead of video data (ablation)SALT student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 77.4 -> 74.9Hurt
teacher pretrained on SSv2 only instead of V-3.6M (ablation)SALT student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 77.4 -> 75.5Hurt
teacher pretrained on K710 only instead of V-3.6M (ablation)SALT student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 77.4 -> 77.4No clear effect
multi-block teacher masking instead of 2x random tube (ablation)SALT student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 76.9 -> 77.4Improved
multi-block teacher masking instead of causal masking (ablation)SALT student · avg of K400/SSv2/IN1k/COINavg top-1 accuracy: 73.7 -> 77.4Improved
shorter teacher training (40k of 240k steps) instead of 80k (ablation)SALT ViT-L · avg top-1 accuracyavg top-1 accuracy: 77.9 -> 78.2Improved
use SALT student loss as model selection signalSALT student · SSv2R2 of loss vs accuracy: R2 = 0.951Improved

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