mlsanityData to 7 Oct 2026

SPEAR NeXT Causal Latent Forecasting Across Multiple Horizons for Spectral Temporal Earth Representation Learning

arXiv 2609.16871 ↗First posted 15 Sep 2026cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Earth observation is inherently dynamic, yet temporal information in many foundation models is learned through reconstruction, invariance, or retrospective sequence summarization. SPEAR NeXT is introduced as a compact pixel-wise multimodal spectral temporal foundation model in which temporal self supervision is formulated as past only, multi horizon latent Earth state prediction. Instantaneous states are first encoded by the pretrained SPEAR model from optical, radar, and environmental observations into compact 32 dimensional embeddings. Their temporal evolution is then modeled by a causally masked Trans former that predicts multiple future latent states from pre ceding observations. Relative temporal order is represented using Rotary Position Embeddings, while month and year embeddings encode seasonal phase and interannual con text.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 23 of its 36 references that are in the map. First posted on arXiv 15 Sep 2026.

Builds on

23 of 36 references.

References · 23

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  2. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  3. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,339
  4. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  5. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 982
  6. 2019Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecastingcited by 748
  7. 2024Revisiting Feature Prediction for Learning Visual Representations from Videocited by 490
  8. 2022SatMAE: Pre-training Transformers for Temporal and Multi-Spectral Satellite Imagerycited by 402
  9. 2022Scale-MAE: A Scale-Aware Masked Autoencoder for Multiscale Geospatial Representation Learningcited by 302
  10. 2023SpectralGPT: Spectral Remote Sensing Foundation Modelcited by 250
  11. 2023SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagerycited by 202
  12. 2023CROMA: Remote Sensing Representations with Contrastive Radar-Optical Masked Autoencoderscited by 167
11 more
  1. 2025AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label datacited by 122
  2. 2023Lightweight, Pre-trained Transformers for Remote Sensing Timeseriescited by 102
  3. 2018Temporal Convolutional Neural Network for the Classification of Satellite Image Time Seriescited by 76
  4. 2024AnySat: One Earth Observation Model for Many Resolutions, Scales, and Modalitiescited by 73
  5. 2019Satellite Image Time Series Classification with Pixel-Set Encoders and Temporal Self-Attentioncited by 71
  6. 2018Multi-Temporal Land Cover Classification with Sequential Recurrent Encoderscited by 61
  7. 2025TESSERA: Temporal Embeddings of Surface Spectra for Earth Representation and Analysiscited by 35
  8. 2021EarthNet2021: A large-scale dataset and challenge for Earth surface forecasting as a guided video prediction taskcited by 34
  9. 2022Generalized Classification of Satellite Image Time Series with Thermal Positional Encodingcited by 14
  10. 2023SICKLE: A Multi-Sensor Satellite Imagery Dataset Annotated with Multiple Key Cropping Parameterscited by 8
  11. 2025Time2Agri: Temporal Pretext Tasks for Agricultural Monitoringcited by 2

Built on by

Not cited yet.