The paper introduces DBCM, a Dual-Branch Calibration Module designed for test-time adaptation in time series forecasting. It addresses non‑stationarity by sharing adaptation across channels via a shared adapter, while adding a low‑rank channel‑specific adapter and channel mixing with gated modulation. Experiments demonstrate that DBCM outperforms existing methods across various datasets, horizons, and backbone models.
arXiv:2610.33822v1 Announce Type: cross Abstract: A central challenge in multivariate forecasting is deciding what to share across channels and where to specialize. Mixture-of-experts models offer fl...
AdaCast introduces a conditional parameter generation framework that tailors time‑series forecasting models to each input by producing low‑rank updates for a frozen pretrained foundation model. Unlike static adaptation methods that apply a single set of updates to all inputs, AdaCast generates input‑specific updates during both training and inference. Experiments on six public benchmarks show that AdaCast consistently outperforms static baselines in in‑domain forecasting and improves zero‑shot generalization to unseen datasets across domains.
arXiv:2606. 10678v1 Announce Type: new Abstract: Transformer-based models have emerged as leading paradigms in time-series forecasting in recent years, employing self-attention mechanisms to capture long-range dependencies.
The paper introduces Chameleon, a channel‑dependent state space model for multivariate time series forecasting that allows data‑dependent, fine‑grained interactions across variables while maintaining linear scaling with the number of variables. By integrating selective state space models with a Kalman filter and adapting GatedDeltaNet as the backbone, Chameleon improves generalization and achieves lower MSE and MAE on strongly dependent ODE and PEMS datasets compared to both channel‑independent and prior channel‑dependent methods. Across 28 benchmark settings, it outperforms baselines in the majority of cases and demonstrates competitive training‑time and memory efficiency on Traffic and ETT datasets.
Aurora‑X is a billion‑parameter time‑series foundation model designed for extreme forecasting tasks. It employs a progressive curriculum that starts with channel‑independent pretraining, then adds cross‑variable dependencies, variable context and horizon lengths, and optional future covariates during mid‑training. A variable‑resolution post‑training stage allows adjustable temporal spans per token at inference, while a pattern‑guided mixture‑of‑experts expands capacity through sparse activation and expert specialization. An implicit quantile network head predicts arbitrary quantiles, enhancing probabilistic forecasting flexibility. Experiments on GIFT‑Eval, TIME, FEV‑Bench, TFB, and DAG‑Bench show state‑of‑the‑art performance against both pretrained TSFMs and task‑specific supervised models.
arXiv:2607. 18899v1 Announce Type: new Abstract: Forecasting under real-world conditions is inherently non-stationary, as the conditional distribution of future observations evolves over time.
arXiv:2606. 28142v1 Announce Type: new Abstract: Test-Time Adaptation (TTA) methods commonly update the affine parameters of normalization layers to adapt deployed models under distribution shifts.
arXiv:2609.15344v1 Announce Type: new Abstract: We study the adaptation of pretrained language models to univariate time-series forecasting through a parameter-efficient transfer learning framework,...
arXiv:2508. 05287v3 Announce Type: replace-cross Abstract: Existing time series foundation models (TSFMs), often based on transformer variants, lack adaptability to different sampling rates, struggle with generalization across varying context and target lengths, and are computationally inefficient.
RATL is a plug‑in method for multivariate time‑series forecasting that uses a frozen base forecaster to build a memory of its historical forecast residuals. During inference, RATL retrieves residual trajectories from similar past contexts and employs a set‑aware router to combine them, providing learned feedback correction. Experiments demonstrate that this residual‑retrieval approach improves the performance of the base forecaster across various benchmarks and backbones.
arXiv:2608. 04051v1 Announce Type: new Abstract: Real-world time series are often governed by recurring patterns, but their dominant periods may vary across datasets, forecasting settings, and individual input windows.
SACQ introduces a structured prediction head for long‑horizon time‑series forecasting that replaces the traditional flatten readout. It uses a two‑stage decoding process: a coarse patch‑grid scaffold is first generated, then each future position is refined via cross‑attention over historical memory, with a learned gate merging the corrections. A batch‑adaptive scaled log‑cosh loss further stabilizes training under long horizons and noisy labels, leading to superior MSE/MAE performance across multiple backbones with minimal overhead.