SPLICE: Latent Diffusion over JEPA Embeddings for Conformal Time-Series Inpainting
Abstract
Long gaps in time series are difficult to reconstruct reliably, and in power systems imputed load values inform dispatch and planning. Existing generative imputers are accurate but provide no finite-sample reliability guarantees under changing conditions. We introduce SPLICE (Self-supervised Predictive Latent Inpainting with Conformal Envelopes), which couples a JEPA encoder, a conditional latent bridge, an hourly-conditioned decoder, and Adaptive Conformal Inference (ACI). Controlled ablations indicate that long-horizon fidelity is driven more by the predictive representation than by generative complexity: under a matched bridge and decoder, JEPA outperforms reconstructive VAE and contrastive TS2Vec encoders, and once the representation is fixed a 5-step flow sampler matches or improves on a 50-step DDIM sampler at a 5-10x speedup. The same ordering is recovered on twelve hydrological catchments. Across thirteen load datasets at 91-day gaps, SPLICE attains the lowest mean Load-only MSE (0.056), winning 9 of 12 non-degenerate datasets against seven established baselines, and the best mean CRPS (0.182, 7.8% below the strongest competitor), while ACI holds 93-95% empirical coverage where static calibration under-covers by up to 7.5 percentage points. A pooled JEPA encoder transfers to four held-out datasets with only bridge and decoder adaptation.
Related
- Clifford BMW Categories and Conformal Embeddings of Types B C and D
- Beyond Imputation: Mask-Adaptive Conformal Prediction via Tree Embeddings on General Missing Data Mechanisms
- Anyon Proliferation and Anyon Superconductivity in Higgsing Transitions via Conformal Embeddings
- The latent structure of diversification from relatedness and network embeddings
- Beyond Prediction: Conformal Inference for Latent Distributional Parameters
BibTeX
@misc{zinflou2026splice,
title = {{SPLICE: Latent Diffusion over JEPA Embeddings for Conformal Time-Series Inpainting}},
author = {Arnaud Zinflou},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2605.00126},
eprint = {2605.00126},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.00126},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - SPLICE: Latent Diffusion over JEPA Embeddings for Conformal Time-Series Inpainting
AU - Arnaud Zinflou
PY - 2026
DA - 2026/10/03
PB - arXiv
DO - 10.48550/arxiv.2605.00126
UR - https://arxiv.org/abs/2605.00126
AB - Long gaps in time series are difficult to reconstruct reliably, and in power systems imputed load values inform dispatch and planning. Existing generative imputers are accurate but provide no finite-sample reliability guarantees under changing conditions. We introduce SPLICE (Self-supervised Predictive Latent Inpainting with Conformal Envelopes), which couples a JEPA encoder, a conditional latent bridge, an hourly-conditioned decoder, and Adaptive Conformal Inference (ACI). Controlled ablations indicate that long-horizon fidelity is driven more by the predictive representation than by generative complexity: under a matched bridge and decoder, JEPA outperforms reconstructive VAE and contrastive TS2Vec encoders, and once the representation is fixed a 5-step flow sampler matches or improves on a 50-step DDIM sampler at a 5-10x speedup. The same ordering is recovered on twelve hydrological catchments. Across thirteen load datasets at 91-day gaps, SPLICE attains the lowest mean Load-only MSE (0.056), winning 9 of 12 non-degenerate datasets against seven established baselines, and the best mean CRPS (0.182, 7.8% below the strongest competitor), while ACI holds 93-95% empirical coverage where static calibration under-covers by up to 7.5 percentage points. A pooled JEPA encoder transfers to four held-out datasets with only bridge and decoder adaptation.
ER -