Safety
No Gaussian Required: Contrastive Inverse Dynamics for JEPA World Models
arXiv:2608.17542v1 Announce Type: cross Abstract: Joint-Embedding Predictive Architectures (JEPAs) learn world models by predicting future embeddings, but the objective admits a trivial solution of a
arXiv:2608.17542v1 Announce Type: cross Abstract: Joint-Embedding Predictive Architectures (JEPAs) learn world models by predicting future embeddings, but the objective admits a trivial solution of a constant encoder, so every practical system adds an anti-collapse mechanism (LeCun, 2022; Assran et al., 2023; Bardes et al., 2022; 2024). LeWorldModel (LeWM) prevents collapse with SIGReg, a regularizer that forces the latent distribution to match an isotropic Gaussian: the representation is stabilized by prescribing what it must look like, independently of the environment it models. We argue that the anti-collapse pressure can instead come from the transition data itself. Action-Contrastive Masked Transition Modeling (AC-MTM) keeps LeWM's forward latent-prediction objective and adds a training-only inverse-dynamics head trained with Action-NCE: each latent transition must identify the action that produced it among the other actions in the batch, a discrimination task that a collapsed encoder provably fails. The inverse branch is discarded after training, leaving test-time encoding, forward prediction, planning, and compute identical to LeWM. On four standard pixel-control tasks under a matched planning protocol, AC-MTM trains stably from scratch and matches SIGReg on average. On the harder multi-object OGBench Visual Scene task, results are consistent with the prescribed geometry becoming a bottleneck: AC-MTM reaches 80.0pm2.0% success versus 58.0pm2.0% for SIGReg, improving by 20-24 points in each training seed. A single 50-episode random-policy run gives a 52% baseline estimate. Contrastive inverse dynamics thus provides a distribution-free anti-collapse signal that requires no target network, stop-gradient, pretrained encoder, or reconstruction objective, and we characterize the action-space and observability assumptions under which it holds. We make our code available at https://github.com/jackboyla/action-contrastive-jepa
Related
- Sub-JEPA: Subspace Gaussian Regularization for Stable End-to-End World Models
- The SIGReg Objective as Variational Free Energy: A Theoretical Active-Inference Account of JEPA World Models
- SJEPA: Learning Elegant Latent Dynamics with Hybrid Symbolic-Neural Predictors
- A Lightweight Library for Energy-Based Joint-Embedding Predictive Architectures
- SCALE: State-Calibrated Latent Embeddings for JEPA Planning in the Right Geometry
Source: arXiv cs.AI | 2026-08-19