AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Var-JEPA: A Variational Formulation of the Joint-Embedding Predictive Architecture – Bridging Predictive and Generative Self-Supervised Learning

Moritz Gögl, Christopher Yau

ICML 2026 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Disentangled and discrete latent models · 0 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
replace deterministic T-JEPA with variational Var-T-JEPAMLP on Covertype · Covertypetest accuracy: 0.408 -> 0.679Improved
replace deterministic T-JEPA with variational Var-T-JEPAMLP on MNIST · MNIST (tabular)test accuracy: 0.113 -> 0.822Improved
add Var-T-JEPA embeddings to raw features baselineMLP on Adult · Adulttest accuracy: 0.849 -> 0.852Improved
use Var-T-JEPA embeddings instead of raw featuresMLP on Covertype · Covertypetest accuracy: 0.750 -> 0.679Hurt
discard 50% most uncertain samples via latent uncertaintyMLP + Var-T-JEPA · Adulttest accuracy: 0.852 -> 0.921Improved
add KL on s_x to ELBO (vs removing it) (ablation)Var-JEPA ELBO · Simulation s_xaggregated KL to N(0,I): 8.374 -> 0.113Improved
add KL terms on all latents (vs removing all) (ablation)Var-JEPA ELBO · Simulation s_xaggregated KL to N(0,I): 10.465 -> 0.113Improved
add reconstruction and generation terms (ablation)Var-JEPA ELBO · Simulation s_xlinear-probe accuracy: 0.571 -> 0.996Improved
add SIGReg to ELBO (ablation)Var-JEPA · Simulation s_xaggregated KL to N(0,I): 0.113 -> 0.107No clear effect
add SIGReg to ELBO without reconstruction and generation (ablation)ELBO (G) · Simulation s_xlinear-probe accuracy: 0.571 -> 0.834Improved
add SIGReg to no-KL ELBO (ablation)ELBO without KL (I) · Simulation s_xaggregated KL to N(0,I): 10.465 -> 2.115Improved
add KL on s_y to ELBO (ablation)Var-JEPA ELBO · Simulation s_ylinear-probe accuracy: 0.983 -> 0.993Improved

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