@_akhaliq: VISReg Variance-Invariance-Sketching Regularization for JEPA training
Summary
Introduces VISReg, a regularization method for JEPA (Joint Embedding Predictive Architecture) training that combines variance, invariance, and sketching constraints.
View Cached Full Text
Cached at: 06/28/26, 07:57 AM
VISReg
Variance-Invariance-Sketching Regularization for JEPA training https://t.co/WFLaqiyzYW
Similar Articles
@reza_byt: World Modeling with JEPA has recently gained traction thanks to a novel anti-collapse mechanism called "SIGReg" (by @yl…
Explains SIGReg, a novel regularizer for JEPA that prevents representation collapse by forcing embeddings to follow an isotropic Gaussian distribution, with theoretical guarantees and a clean training loop.
The SIGReg Objective as Variational Free Energy: A Theoretical Active-Inference Account of JEPA World Models
This paper establishes a formal correspondence between the SIGReg objective used in Joint-Embedding Predictive Architectures (JEPAs) and the Active Inference framework, showing that under certain conditions the training objective becomes an exact variational free energy. It organizes non-contrastive regularisers into an entropy-estimator hierarchy and proves theoretical results including exact information bottleneck and preserved surprise bounds for SIGReg.
The Annotated JEPA
A step-by-step annotated implementation and explanation of Joint Embedding Predictive Architectures (JEPA) for self-supervised learning, covering I-JEPA, V-JEPA, and LeJEPA.
AV-JEPA: Extending LeJEPA to Audio-Visual Self-Supervised Learning
AV-JEPA extends LeJEPA to audio-visual self-supervised learning, achieving cross-modal alignment in latent space without decoders, contrastive negatives, or complex losses, and obtains competitive classification on VGGSound and AudioSet.
Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models
The authors introduce Sub-JEPA, a method using Subspace Gaussian Regularization to improve the stability of end-to-end world models like LeWM, showing consistent performance gains on continuous-control benchmarks.