Morpheus-2.0: Multiscale Probabilistic Forecasting with Dual-Axis Attention
Abstract
A universal forecaster is trained once and applied everywhere, so it must serve tasks whose schemas, forecast horizons, and temporal resolutions all differ. The number of targets and covariates changes from dataset to dataset, horizons vary by orders of magnitude, and sampling rates range from seconds to years. Existing time-series foundation models address these only in part, since they isolate each series or visit the two grid relations in separate blocks, and they fix the horizon and resolution before training begins. We introduce Morpheus-2.0, a compact encoder-only foundation model that represents targets and covariates on a single variable-time grid. Its dual-axis attention runs both relations side by side in every layer, one across variables at a timestep and one across time within a variable; a learned gate weights the two per layer, and an axis-aware rotary encoding preserves both grid coordinates. The forecast length is sampled during training, so one set of weights spans short to long horizons at no additional inference cost. Morpheus-2.0 Multiscale extends the model to several temporal resolutions at once, weighting them according to the series being forecast. Across GIFT-Eval, which spans several domains and frequencies, the two models give complementary operating points: Morpheus-2.0 attains 0.685 relative MASE with 81M parameters, while the larger Morpheus-2.0 Multiscale attains 0.646 relative MASE and 0.455 relative CRPS, the lowest of any published system on both metrics.
Related
- SDECast: Probabilistic Weather Forecasting in Continuous Time with Neural SDEs
- Probabilistic Forecasting of Electric Vehicle Charging in a Time-varying Hierarchical Infrastructure
- Combining Weather Forecast Aggregation and State-Space Models for Adaptive Probabilistic Electricity Load Forecasting
- From morning departures to evening returns: a probabilistic model for forecasting bike-sharing demand
- Rollcast: Proper-Score Gated Rolling Anchors for Adaptive Probabilistic Time-Series Forecasting
BibTeX
@misc{aakriti2026morpheus20,
title = {{Morpheus-2.0: Multiscale Probabilistic Forecasting with Dual-Axis Attention}},
author = {Aakriti and Manoj Cherukumalli and Prathamesh Patil and Leonidas Tsepenekas and Amit Varshney and Penghang Liu and Elizabeth Fons and Vamsi Potluru},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.24321},
eprint = {2610.24321},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.24321},
}
RIS
TY - UNPB
TI - Morpheus-2.0: Multiscale Probabilistic Forecasting with Dual-Axis Attention
AU - Aakriti
AU - Manoj Cherukumalli
AU - Prathamesh Patil
AU - Leonidas Tsepenekas
AU - Amit Varshney
AU - Penghang Liu
AU - Elizabeth Fons
AU - Vamsi Potluru
PY - 2026
DA - 2026/10/16
PB - arXiv
DO - 10.48550/arxiv.2610.24321
UR - https://arxiv.org/abs/2610.24321
AB - A universal forecaster is trained once and applied everywhere, so it must serve tasks whose schemas, forecast horizons, and temporal resolutions all differ. The number of targets and covariates changes from dataset to dataset, horizons vary by orders of magnitude, and sampling rates range from seconds to years. Existing time-series foundation models address these only in part, since they isolate each series or visit the two grid relations in separate blocks, and they fix the horizon and resolution before training begins. We introduce Morpheus-2.0, a compact encoder-only foundation model that represents targets and covariates on a single variable-time grid. Its dual-axis attention runs both relations side by side in every layer, one across variables at a timestep and one across time within a variable; a learned gate weights the two per layer, and an axis-aware rotary encoding preserves both grid coordinates. The forecast length is sampled during training, so one set of weights spans short to long horizons at no additional inference cost. Morpheus-2.0 Multiscale extends the model to several temporal resolutions at once, weighting them according to the series being forecast. Across GIFT-Eval, which spans several domains and frequencies, the two models give complementary operating points: Morpheus-2.0 attains 0.685 relative MASE with 81M parameters, while the larger Morpheus-2.0 Multiscale attains 0.646 relative MASE and 0.455 relative CRPS, the lowest of any published system on both metrics.
ER -