SC-JEPA: Stabilizing Latent Predictive Learning for Time-Series Anomaly Prediction

Extended and revised version of MTS-JEPA (arXiv:2602.04643).

[Paper] [arXiv]

Abstract

Time-series anomaly prediction aims to forecast future system failures before they emerge, making latent predictive models such as JEPA a promising framework for capturing precursor dynamics. However, directly applying continuous self-distillation to time-series data is often unstable and can lead to representation collapse, while also struggling to model precursors evolving at different temporal scales. To address this, we propose SC-JEPA, a new JEPA-based framework to model time-series anomaly prediction in a discretized predictive state space. It introduces a soft codebook bottleneck to stabilize latent predictive learning and encourage regime-level structure in the learned representations. Building on this stabilized latent space, we further design a multi-resolution predictive objective to capture precursor patterns at different temporal scales. Experiments on five benchmarks show that SC-JEPA achieves strong early-warning performance.

Main Contributions

• Soft Codebook Bottleneck: Introduces differentiable quantization into the JEPA latent space to prevent representation collapse and impose regime-level discrete structure.
• Multi-Resolution Predictive Objective: Captures precursor patterns across temporal scales, from sudden shocks to slow drifts, within a single stabilized latent space.
• Discretized Predictive State Space: Reformulates time-series anomaly prediction as latent prediction in a discrete state space, improving both stability and early-warning sensitivity.

BibTeX

@inproceedings{
he2026scjepa,
title={{SC}-{JEPA}: Stabilizing Latent Predictive Learning for Time-Series Anomaly Prediction},
author={Yanan He and Yunshi Wen and Xin Wang and Tengfei Ma},
booktitle={Forecasting as a New Frontier of Intelligence},
year={2026},
url={https://openreview.net/forum?id=SNjhsnUvhH}
}