# Is the reconstruction loss culprit? An attempt to outperform JEPA

arXiv: https://arxiv.org/abs/2603.14131 · HTML: https://mlsanity.com/paper/2603.14131/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 3 of its 4 references that are in the map.
- First posted on arXiv 14 Mar 2026.

## Builds on (most cited first, 3 of 3)

- [World Models](https://mlsanity.com/paper/1803.10122/) (2018)
- [WoVR: World Models as Reliable Simulators for Post-Training VLA Policies with RL](https://mlsanity.com/paper/2602.13977/) (2026)
- [A Lightweight Library for Energy-Based Joint-Embedding Predictive Architectures](https://mlsanity.com/paper/2602.03604/) (2026)

## Built on by (most cited first, 1 of 1)

- [Does Joint-Embedding Predictive Architecture Pretraining Help Time Series Forecasting?](https://mlsanity.com/paper/2609.31680/) (2026)

## Abstract (arXiv)

We evaluate JEPA-style predictive representation learning versus reconstruction-based autoencoders on a controlled "TV-series" linear dynamical system with known latent state and a single noise parameter. While an initial comparison suggests JEPA is markedly more robust to noise, further diagnostics show that autoencoder failures are strongly influenced by asymmetries in objectives and by bottleneck/component-selection effects (confirmed by PCA baselines). Motivated by these findings, we introduce gated predictive autoencoders that learn to select predictable components, mimicking the beneficial feature-selection behavior observed in over-parameterized PCA. On this toy testbed, the proposed gated model is stable across noise levels and matches or outperforms JEPA.