Scale-Invariant Training for Time Series Foundation Models

arXiv:2610.07324 · cs.LG, cs.AI · Submitted 2026-10-05 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-10-05

Updated: 2026-10-09

Code: https://github.com/Nixtla/statsforecast

Terminology

• Foundation Model — 1× in this paper · explained in 3 episodes
• Foundation Models — 1× in this paper · explained in 7 episodes

Sources

• Chronos-2: From Univariate to Universal Forecasting
• An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
• On the Role of Reversible Instance Normalization
• Kairos: Toward Adaptive and Parameter-Efficient Time Series Foundation Models
• Toto 2.0: Time Series Forecasting Enters the Scaling Era
• Adam: A Method for Stochastic Optimization
• Moirai 2.0: When Less Is More for Time Series Forecasting
• Timer-S1: A Billion-Scale Time Series Foundation Model with Serial Scaling
• Cross-Frequency Time Series Meta-Forecasting
• Does Normalization Choice Matter for Causal Large Time-Series Models?
• Output Scaling: YingLong-Delayed Chain of Thought in a Large Pretrained Time Series Forecasting Model
• Into the ORBIT for Time Series: Training Regimes for Foundation Models

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks