Time-Series Foundation Models

Time Series Foundation Models (TSFMs) excel at numerical forecasting but operate as black boxes lacking qualitative reasoning. Conversely, applying LLMs directly to temporal data introduces a modality gap: text tokenizers fragment continuous numerical values, degrading mathematical relationships and exploding sequence lengths, leading to computational overhead. To resolve this, we introduce STRIDE (Strategic Time-series Reasoning Injected via Distilled Embeddings), a novel framework natively integrating LLM reasoning into the continuous embedding space of TSFMs. Instead of discrete tokens, STRIDE distills reasoning traces into a lightweight LLM and dynamically projects its pooled hidden states as a continuous cross-modal prior into the numerical encoder. Jointly optimized via cross-entropy and quantile losses, STRIDE readily attaches to diverse architectures (e.g., Chronos-2, Timer-S1). STRIDE ranks #1 on the GIFT-Eval leaderboard (0.625 MASE, 0.423 CRPS) when ensembled with Synapse. On TFRBench, STRIDE consistently outperforms baselines in both forecasting accuracy and reasoning quality, with ablations confirming gains stem from strategic reasoning. Thus, injecting semantic reasoning as a continuous prior equips TSFMs with human-interpretable reasoning while improving predictive accuracy.

Using Chronos-2, an open-source time-series foundation model, we evaluate pretrained time-series models for economic and financial forecasting with an emphasis on whether multivariate (MV) inputs improve accuracy relative to univariate (UV) baselines. The study covers two panels -- the Magnificent-7 equities and U.S. Treasury interest rates -- as well as a combined panel, using rolling monthly evaluations from 2000--2025. We vary input window lengths and forecast horizons and report RMSE and MAPE. Across datasets, MV forecasts consistently outperform UV forecasts, with especially strong gains for interest rates and meaningful improvements for equities. Series-level comparisons show MV improvements in every case, and error dispersion is generally lower under MV inputs. We also provide parameter-heatmap and time-series visualizations. However, mixing time series across equity and interest rate markets reduces forecast accuracy, indicating that adding noisy context degrades model performance. Overall, the results indicate that foundation models can leverage cross-series information to improve forecast accuracy in finance, and that the benefits are strongest when related series are modeled jointly under disciplined rolling protocols. Other than using an open-source foundation model, this paper also showcases how AI may be used for financial research.

Health foundation models (FMs) learn useful representations from wearable sensors, but interpreting what they encode and transferring that knowledge across modalities after training remains difficult. We present a post-training framework that decomposes frozen embeddings into interpretable directions, referred to as symbols, and use these symbols to align the embedding spaces without retraining. We evaluate the framework on three FMs for photoplethysmography (PPG) and accelerometer data, independently pretrained on ~20M minutes of unlabeled data from ~172K participants, and analyzed on a held-out cohort of 30K subjects. We find that extracted symbols associate selectively with health conditions and physiological attributes, and these associations are partially shared across modalities and architectures. Cross-modal transfer via symbols retains more than 95% of in-domain performance, is nearly symmetric across domain directions, and saturates with limited paired data, together indicating that alignment recovers a shared low-dimensional subspace rich in physiological information. Overall, these results suggest that health FM embeddings contain an interpretable symbolic organization that is shared across modalities and supports cross-domain transfer without joint training.

Scientific foundation models are expected to reuse representations under changes in dataset, acquisition protocol, and deployment domain, yet many sequence backbones treat scientific temporal structure as an unconstrained pattern to be fitted. We argue that this misses a central property of natural dynamical systems: neural and atmospheric time series are organized by interacting processes across multiple physical timescales, and failure to preserve this multiscale structure contributes to brittleness under distribution shift. We formalize this failure mode as temporal kernel mismatch, where a model fits in-distribution dynamics with an effective memory policy that is not anchored to the signal's physical timescales, leading to representation drift and degraded transfer. We propose Physics-Informed Multi-Scale Mamba (PIMSM), a state-space architecture that maps spectrum-estimated transition points between frequency regimes (knee frequencies) to scale-specific discretization parameters and anchors them to acquisition time units. On Human Connectome Project fMRI, PIMSM improves robustness and representation stability under severe temporal-context truncation, extreme low-resource transfer, and resting-state-to-task-state generalization. Without modality-specific adaptation, the same architecture also attains the lowest variable-wise MAE across all reported horizons and variables on Weather-5K held-out-station spatial out-of-distribution forecasting. These results support temporal-scale alignment as a practical inductive bias for scientific foundation models that must preserve structure, not only fit correlations, under deployment shift.

This paper presents a preliminary analysis of the ability of Chronos foundation model to process and internally represent frequency domain information. Foundation models that process time-series data offer practitioners a unified architecture capable of learning generic temporal representations across diverse tasks and domains, reducing the need for task-specific feature engineering and enabling transfer across signal modalities. Despite their growing adoption, the extent to which such models encode fundamental signal properties remains insufficiently characterised. We address this gap by analysing Chronos under controlled conditions, starting from the simplest class of signals: discrete sinusoids generated at fixed frequencies. Using lightweight online minimum description length probes applied to the decoder architecture, we test for the presence and separability of frequency information in the model's internal representations. The results provide insight into how frequential content is captured across the frequency spectrum and highlight regimes in which representation quality may degrade or require particular care. These findings offer practical guidance for users of Chronos in signal processing and information fusion contexts, and contribute to ongoing efforts to improve the interpretability and evaluation of foundation models for temporal data.

Time Series Foundation Models (TSFMs) have borrowed the long context paradigm from natural language processing under the premise that feeding more history into the model improves forecast quality. But in stochastic domains, distant history is often just high-frequency noise, not signal. Hence, the proposed work tests whether this premise actually holds by running continuous context architectures (PatchTST included) through the ETTh1 benchmark. The obtained results contradict the premise: an inverse scaling law shows up clearly, with forecasting error rising as context gets longer. A 3,000-step window causes performance to drop by over 68%, evidence that attention mechanisms are poor at ignoring irrelevant historical volatility. Retrieval-Augmented Forecasting (RAFT) is evaluated as an alternative. RAFT achieves a mean squared error (MSE) of 0.379 with a fixed 720-step window and selective retrieval, outperforming both long-context configurations and zero-shot foundation models (Chronos, Moirai) despite requiring far less computation. In addition, the retrieval step injects only the most relevant historical segments as dynamic exogenous variables, which gives the model a context-informed inductive bias it cannot build on its own from raw sequences. Therefore, foundation models going forward need to shift architecturally toward selective retrieval.

Time Series Foundation Models (TSFMs) have recently emerged as general-purpose forecasting models and show considerable potential for applications in energy systems. However, applications in critical infrastructure like power grids require transparency to ensure trust and reliability and cannot rely on pure black-box models. To enhance the transparency of TSFMs, we propose an efficient algorithm for computing Shapley Additive Explanations (SHAP) tailored to these models. The proposed approach leverages the flexibility of TSFMs with respect to input context length and provided covariates. This property enables efficient temporal and covariate masking (selectively withholding inputs), allowing for a scalable explanation of model predictions using SHAP. We evaluate two TSFMs - Chronos-2 and TabPFN-TS - on a day-ahead load forecasting task for a transmission system operator (TSO). In a zero-shot setting, both models achieve predictive performance competitive with a Transformer model trained specifically on multiple years of TSO data. The explanations obtained through our proposed approach align with established domain knowledge, particularly as the TSFMs appropriately use weather and calendar information for load prediction. Overall, we demonstrate that TSFMs can serve as transparent and reliable tools for operational energy forecasting.

Transthoracic echocardiography is the reference standard for confirming structural heart disease (SHD), but first-line screening is limited by cost, workflow burden, and specialist availability. We evaluated whether open pretrained electrocardiogram (ECG) foundation models can support echo-confirmed multi-label SHD detection using the public EchoNext Mini-Model benchmark. Six echocardiography-derived abnormalities were targeted: reduced left ventricular ejection fraction, increased left ventricular wall thickness, aortic stenosis, mitral regurgitation, tricuspid regurgitation, and right ventricular systolic dysfunction. Under a common pipeline, we compared engineered ECG features with gradient boosting, end-to-end waveform learning from scratch, and transfer from open ECG foundation models. We then applied in-domain self-supervised adaptation of an ECG foundation model (ECG-FM) on EchoNext waveforms followed by selective supervised fine-tuning, and evaluated trade-offs between discrimination and adaptation cost. Adapted ECG-FM models achieved the best overall performance: peak macro-AUROC 0.8509 and macro-AUPRC 0.4297, while a parameter-efficient operating point preserved AUROC (0.8501) and attained the highest fixed-threshold macro-F1 0.3691. Late fusion with covariates did not improve threshold-independent discrimination, and evaluated LoRA, alternative backbones, and mixture-of-foundations strategies did not surpass the best adapted single-backbone models. These results indicate that for ECG-based case finding and echocardiography triage, combining target-domain self-supervised adaptation with selective supervised updating of a pretrained ECG backbone is the most effective transfer strategy.

Driven by the transition towards a climate-neutral energy system, accurate energy time series forecasting is critical for planning and operation. Yet, it remains largely a dataset-specific task, requiring comprehensive training data, limiting scalability, and resulting in high model development and maintenance effort. Recently, foundation models that aim to learn generalizable patterns via extensive pretraining have shown superior performance in multiple prediction tasks. Despite their success and strong potential to address challenges in energy forecasting, their application in this domain remains largely unexplored. We address this gap by presenting the Foundation Models in Energy Time Series Forecasting (FETS) benchmark. We (1) provide a structured overview of energy forecasting use cases along three main dimensions: stakeholders, attributes, and data categories; (2) collect and analyze 54 datasets across 9 data categories, guided by typical stakeholder interests; (3) benchmark foundation models against classical machine learning approaches across different forecasting settings. Foundation models consistently outperform dataset-specific optimized machine learning approaches across all settings and data categories, despite the latter having seen the full historic target data during training. In particular, covariate-informed foundation models achieve the strongest performance. Further analysis reveals a strong correlation between predictive performance and spectral entropy, performance saturation beyond a certain context length, and improved performance at higher aggregation levels such as national load, district heating, and power grid data. Overall, our findings highlight the strong potential of foundation models as scalable and generalizable forecasting solutions for the energy domain, particularly in data-constrained and privacy-sensitive settings.

Transformer-based scientific foundation models are increasingly deployed in high-stakes settings, but current architectures give deterministic outputs and provide limited support for calibrated predictive uncertainty. We propose Stochastic Attention, a sample average lightweight inference-time modification that randomizes attention by replacing softmax weights with normalized multinomial samples controlled by a single concentration parameter, and produces predictive ensembles without retraining. To set this parameter, we introduce a calibration objective that matches the stochastic attention output with the target, yielding an efficient univariate post-hoc tuning problem. We evaluate this mechanism on scientific foundation models for weather and time-series forecasting, as well as several regression tasks. Across benchmarks against uncertainty-aware baselines, we find that Sample Average Stochastic Attention achieves the strongest native calibration and the sharpest prediction intervals at comparable calibration, with adaptation costs nearly three orders of magnitude lower than the next-best baseline.

Non-stationary sequences arise naturally in control, forecasting, and decision-making. The data-generating process shifts at unknown times, and models must detect the change, discard or downweight obsolete evidence, and adapt to new dynamics on the fly. Transformer-based foundation models increasingly rely on in-context learning for time series forecasting, tabular prediction, and continuous control. As these models are deployed in non-stationary environments, understanding their ability to detect and adapt to regime shifts is important. We formalize this as an in-context change-point detection problem and formally establish the existence of transformer models that solve this problem. Our construction demonstrates that model complexity, in layers and parameters, depends on the level of information available about the change-point location, from no knowledge to knowing exact timing. We validate our results with experiments on synthetic linear regression and linear dynamical systems, where trained transformers match the performance of optimal baselines across information levels. We also show that encoding and incorporating changepoint knowledge indeed improves the real-world performance of a pretrained foundation models on infectious disease forecasting and on financial volatility forecasting around Federal Open Market Committee (FOMC) announcements without retraining, demonstrating practical applicability to real-world regime changes.

Time Series Foundation Models (TSFMs) advance generalization and data efficiency in time series forecasting by unified large-scale pretraining. But TSFMs remain lacking when adapting to specific downstream forecasting tasks for two reasons. First, the non-stationary and uncertain nature of time series data lead to inevitable temporal distribution shifts between historical training and future testing data, while current Supervised FineTuning (SFT)-based methods are prone to overfitting and may degrade generalization. Second, training data availability varies across forecasting tasks, requiring TSFMs to generalize well under diverse data regimes. To address these challenges, we introduce the Time series Reinforcement Finetuning (TimeRFT) paradigm for TSFM downstream adaptation, which consists of two task-specific training recipes: i) A forecasting quality-based temporal reward mechanism that conducts a multi-faceted evaluation of the contribution of each prediction step to overall forecasting accuracy. ii) A forecasting difficulty-based data selection strategy to identify time series samples with generalizable predictive patterns and informative training signals. Extensive experiments demonstrate TimeRFT can consistently outperform SFT-based adaptation methods across various real-world forecasting tasks and training data regimes, enhancing prediction accuracy and generalization against unforeseen distribution shifts.

Oil and gas drilling operations generate extensive time-series data from surface sensors, yet accurate real-time prediction of critical downhole metrics remains challenging due to the scarcity of labelled downhole measurements. This systematic mapping study reviews thirteen papers published between 2015 and 2025 to assess the potential of Masked Autoencoder Foundation Models (MAEFMs) for predicting downhole metrics from surface drilling data. The review identifies eight commonly collected surface metrics and seven target downhole metrics. Current approaches predominantly employ neural network architectures such as artificial neural networks (ANNs) and long short-term memory (LSTM) networks, yet no studies have explored MAEFMs despite their demonstrated effectiveness in time-series modeling. MAEFMs offer distinct advantages through self-supervised pre-training on abundant unlabeled data, enabling multi-task prediction and improved generalization across wells. This research establishes that MAEFMs represent a technically feasible but unexplored opportunity for drilling analytics, recommending future empirical validation of their performance against existing models and exploration of their broader applicability in oil and gas operations.

As power systems transition toward renewable-rich and inverter-dominated operations, accurate time-domain dynamic analysis becomes increasingly critical. Such analysis supports key operational tasks, including transient stability assessment, dynamic security analysis, contingency screening, and post-fault trajectory evaluation. In practice, these tasks may operate under several challenges, including unknown and time-varying system parameters, privacy constraints on data sharing, and the need for fast online inference. Existing learning-based approaches are typically trained for individual systems and therefore lack generalization across operating conditions and physical parameters. Hence, this paper proposes LArge Scale Small ODE (LASS)-ODE-Power, a learning framework for general-purpose time-domain prediction. The proposed approach leverages large-scale pretraining on more than 40 GB of DAE or ordinary differential-equation (ODE) trajectories to learn transferable representations. The resulting model supports trajectory prediction from short measurement prefixes across diverse dynamic regimes, including electromechanical and inverter-driven systems. Hence, the model can be directly used without data sharing in a zero-shot setting. In addition, the proposed architecture incorporates parallel and linearized computation to achieve fast inference. Moreover, to enhance task-specific performance in power systems, a specialized fine-tuning strategy is developed based on approximately 1 GB of heterogeneous power-system dynamic data. Extensive experiments over diverse power-system simulation scenarios demonstrate that LASS-ODE-Power consistently outperforms existing learning-based models in trajectory prediction accuracy with efficient inference.

Large-scale renewable energy deployment introduces pronounced volatility into the electricity system, turning grid operation into a complex stochastic optimization problem. Accurate electricity price forecasting (EPF) is essential not only to support operational decisions, such as optimal bidding strategies and balancing power preparation, but also to reduce economic risk and improve market efficiency. Probabilistic forecasts are particularly valuable because they quantify uncertainty stemming from renewable intermittency, market coupling, and regulatory changes, enabling market participants to make informed decisions that minimize losses and optimize expected revenues. However, it remains an open question which models to employ to produce accurate forecasts. Should these be task-specific machine learning (ML) models or Time Series Foundation Models (TSFMs)? In this work, we compare four models for day-ahead probabilistic EPF (PEPF) in European bidding zones: a deterministic NHITS backbone with Quantile-Regression Averaging (NHITS+QRA) and a conditional Normalizing-Flow forecaster (NF) are compared with two TSFMs, namely Moirai and ChronosX. On the one hand, we find that TSFMs outperform task-specific deep learning models trained from scratch in terms of CRPS, Energy Score, and predictive interval calibration across market conditions. On the other hand, we find that well-configured task-specific models, particularly NHITS combined with QRA, achieve performance very close to TSFMs, and in some scenarios, such as when supplied with additional informative feature groups or adapted via few-shot learning from other European markets, they can even surpass TSFMs. Overall, our findings show that while TSFMs offer expressive modeling capabilities, conventional models remain highly competitive, emphasizing the need to weigh computational expense against marginal performance improvements in PEPF.

Current fMRI foundation models primarily rely on a limited range of brain states and mismatched pretraining tasks, restricting their ability to learn generalized representations across diverse brain states. We present Brain-DiT, a universal multi-state fMRI foundation model pretrained on 349,898 sessions from 24 datasets spanning resting, task, naturalistic, disease, and sleep states. Unlike prior fMRI foundation models that rely on masked reconstruction in the raw-signal space or a latent space, Brain-DiT adopts metadata-conditioned diffusion pretraining with a Diffusion Transformer (DiT), enabling the model to learn multi-scale representations that capture both fine-grained functional structure and global semantics. Across extensive evaluations and ablations on 7 downstream tasks, we find consistent evidence that diffusion-based generative pretraining is a stronger proxy than reconstruction or alignment, with metadata-conditioned pretraining further improving downstream performance by disentangling intrinsic neural dynamics from population-level variability. We also observe that downstream tasks exhibit distinct preferences for representational scale: ADNI classification benefits more from global semantic representations, whereas age/sex prediction comparatively relies more on fine-grained local structure.

Foundation models have transformed natural language processing and computer vision, and a rapidly growing literature on time-series foundation models (TSFMs) seeks to replicate this success in forecasting. While recent open-source models demonstrate the promise of TSFMs, the field lacks a comprehensive and community-accepted model evaluation framework. We see at least four major issues impeding progress on the development of such a framework. First, existing evaluation frameworks comprise benchmark forecasting tasks derived from often outdated datasets (e.g., M3), many of which lack clear metadata and overlap with the corpora used to pre-train TSFMs. Second, these frameworks evaluate models along a narrowly defined set of benchmark forecasting tasks, such as forecast horizon length or domain, but overlook core statistical properties such as non-stationarity and seasonality. Third, domain-specific models (e.g., XGBoost) are often compared unfairly, as existing frameworks do not enforce a systematic and consistent hyperparameter tuning convention for all models. Fourth, visualization tools for interpreting comparative performance are lacking. To address these issues, we introduce TempusBench, an open-source evaluation framework for TSFMs. TempusBench consists of 1) new datasets which are not included in existing TSFM pretraining corpora, 2) a set of novel benchmark tasks that go beyond existing ones, 3) a model evaluation pipeline with a standardized hyperparameter tuning protocol, and 4) a tensorboard-based visualization interface. We provide access to our code on GitHub: https://github.com/Smlcrm/TempusBench and maintain a live leaderboard at https://smlcrm.com/tempusbench.

Time series foundation models (TSFMs) have recently achieved remarkable success in universal forecasting by leveraging large-scale pretraining on diverse time series data. Complementing this progress, incorporating frequency-domain information yields promising performance in enhancing the modeling of complex temporal patterns, such as periodicity and localized high-frequency dynamics, which are prevalent in real-world time series. To advance this direction, we propose a new perspective that integrates explicit frequency-domain representations into scalable foundation models, and introduce WaveMoE, a wavelet-enhanced mixture-of-experts foundation model for time series forecasting. WaveMoE adopts a dual-path architecture that jointly processes time series tokens and wavelet tokens aligned along a unified temporal axis, and coordinates them through a shared expert routing mechanism that enables consistent expert specialization while efficiently scaling model capacity. Preliminary experimental results on 16 diverse benchmark datasets indicate that WaveMoE has the potential to further improve forecasting performance by incorporating wavelet-domain corpora.

Recent work on time-series models has leveraged self-supervised training to learn meaningful features and patterns in order to improve performance on downstream tasks and generalize to unseen modalities. While these pretraining methods have shown great promise in one-to-many scenarios, where a model is pre-trained on one dataset and fine-tuned on a downstream dataset, they have struggled to generalize to new datasets when more datasets are added during pre-training. This is a fundamental challenge in building foundation models for time-series data, as it limits the ability to develop models that can learn from a large variety of diverse datasets available. To address this challenge, we present a new pre-training paradigm for time-series data called ADAPT, which can efficiently align the physical properties of data in the time-series domain, enabling mixed-batch pre-training despite the extreme discrepancies in the input sizes and channel dimensions of pre-training data. We trained on 162 time-series classification datasets and set new state-of-the-art performance for classification benchmarks. We successfully train a model within the time-series domain on a wide range of datasets simultaneously, which is a major building block for building generalist foundation models in time-series domains.

Heterogeneity in time series data is more pronounced than in vision or language, as temporal dynamics vary substantially across domains and tasks. Existing efforts on training time series foundation models (TSFMs) from scratch are often trained with mixed-batch strategies that merge large-scale datasets, which can cause gradient conflicts and degrade representation quality. To address this, we propose a fine-grained learning method that distills invariant knowledge from heterogeneous series while reducing cross-domain interference. We characterize heterogeneity at two levels: inter-domain and intra-domain. To tackle this bi-level heterogeneity, we design a federated learning method that mitigates intra-domain conflicts by enforcing domain-invariant and semantically consistent representations through local regularization, and addresses inter-domain discrepancies by enhancing cross-domain collaboration via domain-aware aggregation. Experiments across diverse benchmarks show that TSFMs trained with our method consistently outperform both centralized and federated TSFM baselines in point and probabilistic forecasting, while also achieving competitive zero-shot performance at scale, offering a flexible pathway for training TSFMs from scratch in heterogeneous environments.

Data-driven models for building thermal dynamics are a scalable approach for enabling energy-efficient operation through fault detection & diagnosis or advanced control. To obtain accurate models, measurement data from a target building spanning months to years are required. Transfer Learning (TL) mitigates this challenge by employing pretrained models based on single or multiple source buildings. General multi-source TL models promise to outperform single-source TL, but alternative multi-source modeling architectures remain to be explored, and evaluation on real-world data is missing. Moreover, time series foundation models (TSFM) have emerged as candidates for the best-performing general models. Hence, we conduct a first, comprehensive assessment of general modeling approaches for building thermal dynamics, including multi-source TL and TSFMs. Our assessment includes ablations using four state-of-the-art multi-source TL architectures and evaluations on synthetic as well as real-world data. We demonstrate that multi-source TL models are highly effective in accurately modeling buildings in real-world applications, yielding up to 63% lower forecasting errors compared to single-source TL. Moreover, our results suggest a trade-off between multi-source TL models exclusively pretrained with building data and TSFMs pretrained with a multitude of different time series, revealing that data from 16-32 source buildings must be available over 1 year for pretraining multi-source TL models to consistently outperform TSFMs as evaluated using the mean absolute error. These findings provide practical guidance for selecting modeling strategies based on the number of source buildings available for pretraining multi-source TL models.

Leveraging Large Language Models (LLMs) as federated learning (FL)-based time series foundation models offers a promising way to transfer the generalization capabilities of LLMs to time series data while preserving access to private data. However, the semantic misalignment between time-series data and the text-centric latent space of existing LLMs often leads to degraded performance. Meanwhile, the parameter-sharing mechanism in existing FL methods model heterogeneous cross-domain time-series data into a unified continuous latent space, which contradicts the fact that time-series semantics frequently manifest as discrete and recurring regimes. To address these limitations, we propose FeDPM, a federated framework for time-series foundation models based on discrete prototypical memories. Specifically, we learn local prototypical memory priors for intra-domain time-series data. We then align cross-domain memories to promote a unified discrete latent space and introduce a domain-specific memory update mechanism to balance shared and personalized prototypical knowledge. Extensive experiments demonstrate the efficiency and effectiveness of FeDPM. The code is publicly available at https://anonymous.4open.science/r/FedUnit-64D1.

Time series foundation models (TSFMs) are widely used as generic feature extractors, yet the notion of non-stationarity in their embedding spaces remains poorly understood. Recent work often conflates non-stationarity with distribution shift, blurring distinctions fundamental to classical time-series analysis and long-standing methodologies such as statistical process control (SPC). In SPC, non-stationarity signals a process leaving a stable regime - via shifts in mean, variance, or emerging trends - and detecting such departures is central to quality monitoring and change-point analysis. Motivated by this diagnostic tradition, we study how different forms of distributional non-stationarity - mean shifts, variance changes, and linear trends - become linearly accessible in TSFM embedding spaces under controlled conditions. We further examine temporal non-stationarity arising from persistence, which reflects violations of weak stationarity due to long-memory or near-unit-root behavior rather than explicit distributional shifts. By sweeping shift strength and probing multiple TSFMs, we find that embedding-space detectability of non-stationarity degrades smoothly and that different models exhibit distinct, model-specific failure modes.

The Radio Access Network (RAN) is evolving into a programmable and disaggregated infrastructure that increasingly relies on AI-native algorithms for optimization and closed-loop control. However, current RAN intelligence is still largely built from task-specific models tailored to individual functions, resulting in model fragmentation, limited knowledge sharing across tasks, poor generalization, and increased system complexity. To address these limitations, we introduce TimeRAN, a unified multi-task learning framework for time-series modeling in the RAN. TimeRAN leverages a lightweight time-series foundation model with few task-specific heads to learn transferable representations that can be efficiently adapted across diverse tasks with limited supervision. To enable large-scale pretraining, we further curate and open-source TimeRAN DataPile, the largest time-series corpus for RAN analytics to date, comprising over 355K time series and 0.56B measurements across diverse telemetry sources, protocol layers, and deployment scenarios. We evaluate TimeRAN across a comprehensive set of RAN analytics tasks, including anomaly detection, classification, forecasting, and imputation, and show that it achieves state-of-the-art performance with minimal or no task-specific fine-tuning. Finally, we integrate TimeRAN into a proof-of-concept 5G testbed and demonstrate that it operates efficiently with limited resource requirements in real-world scenarios.

Recent work suggests that large-scale, multi-animal modeling can significantly improve neural recording analysis. However, for functional calcium traces, existing approaches remain task-specific, limiting transfer across common neuroscience objectives. To address this challenge, we propose CalM, a self-supervised neural foundation model trained solely on neuronal calcium traces and adaptable to multiple downstream tasks, including forecasting and decoding. Our key contribution is a pretraining framework, composed of a high-performance tokenizer mapping single-neuron traces into a shared discrete vocabulary, and a dual-axis autoregressive transformer modeling dependencies along both the neural and the temporal axis. We evaluate CalM on a large-scale, multi-animal, multi-session dataset. On the neural population dynamics forecasting task, CalM achieves competitive performance against strong specialized baselines after pretraining. With a task-specific head, CalM further adapts to the behavior decoding task and achieves superior results compared with supervised decoding models. Moreover, linear analyses of CalM representations reveal interpretable functional structures beyond predictive accuracy. Taken together, we propose a novel and effective self-supervised pretraining paradigm for foundation models based on calcium traces, paving the way for scalable pretraining and broad applications in functional neural analysis. Code is released at https://github.com/TSuXinH/CalM.

Time series forecasting is critical across finance, healthcare, and cloud computing, yet progress is constrained by a fundamental bottleneck: the scarcity of large-scale, high-quality benchmarks. To address this gap, we introduce QuitoBench, a regime-balanced benchmark for time series forecasting with coverage across eight trend$\times$seasonality$\times$forecastability (TSF) regimes, designed to capture forecasting-relevant properties rather than application-defined domain labels. The benchmark is built upon Quito, a billion-scale time series corpus of application traffic from Alipay spanning nine business domains. Benchmarking 10 models from deep learning, foundation models, and statistical baselines across 232,200 evaluation instances, we report four key findings: (i) a context-length crossover where deep learning models lead at short context ($L=96$) but foundation models dominate at long context ($L \ge 576$); (ii) forecastability is the dominant difficulty driver, producing a $3.64 \times$ MAE gap across regimes; (iii) deep learning models match or surpass foundation models at $59 \times$ fewer parameters; and (iv) scaling the amount of training data provides substantially greater benefit than scaling model size for both model families. These findings are validated by strong cross-benchmark and cross-metric consistency. Our open-source release enables reproducible, regime-aware evaluation for time series forecasting research.

In-context learning (ICL) enables task adaptation at inference time by conditioning on demonstrations rather than updating model parameters. Although recent time-series foundation models incorporate contextual conditioning, retrieval, or example-based prompting, they typically rely on implicit positional structure or task-specific objectives rather than explicit instruction-conditioned input-output demonstrations. We introduce iAmTime, a time-series foundation model trained with instruction-conditioned amortized meta-learning to infer tasks directly from example demonstrations. iAmTime represents each episode as a structured prompt over historical context and future-known variables using specialized semantic tokens that attend to designated time-series regions, exchange information across demonstrations, and inject task information into the query representation. The model combines a Hierarchical Multi-Scope Transformer Encoder, which captures temporal and covariate dynamics while inferring latent task structure from demonstrated input-output mappings, with a Task-Conditioned Patch Decoder, which adapts decoding through expert-based routing. We train iAmTime on large-scale real and synthetic corpora using supervised and self-supervised instruction-conditioned tasks, including forecasting, imputation, reconstruction, classification, anomaly detection, and source de-mixing. Across diverse domains, frequencies, and horizons, iAmTime improves zero-shot adaptation over strong time-series foundation baselines on probabilistic and point forecasting benchmarks, while achieving competitive or superior performance on four non-forecasting tasks.

Most existing Time Series Foundation Models (TSFMs) use channel independent modeling and focus on capturing and generalizing temporal dependencies, while neglecting the correlations among channels or overlooking the different aspects of correlations. However, these correlations play a vital role in Multivariate time series forecasting. To address this, we propose a CoRrelation-aware Adapter (CoRA), a lightweight plug-and-play method that requires only fine-tuning with TSFMs and is able to capture different types of correlations, so as to improve forecast performance. Specifically, to reduce complexity, we innovatively decompose the correlation matrix into low-rank Time-Varying and Time-Invariant components. For the Time-Varying component, we further design learnable polynomials to learn dynamic correlations by capturing trends or periodic patterns. To learn positive and negative correlations that appear only among some channels, we introduce a novel dual contrastive learning method that identifies correlations through projection layers, regulated by a Heterogeneous-Partial contrastive loss during training, without introducing additional complexity in the inference stage. Extensive experiments on 10 real-world datasets demonstrate that CoRA can improve TSFMs in multivariate forecasting performance.

Scientific time series are central to scientific AI but are typically sparse, highly heterogeneous, and limited in scale, making unified representation learning particularly challenging. Meanwhile, foundation models pretrained on relevant time series domains such as audio, general time series, and brain signals contain rich knowledge, but their applicability to scientific signals remains underexplored. In this paper, we investigate the transferability and complementarity of foundation models from relevant time series domains, and study how to effectively leverage them to build a unified encoder for scientific time series. We first systematically evaluate relevant foundation models, showing the effectiveness of knowledge transfer to scientific tasks and their complementary strengths. Based on this observation, we propose STEP, a Scientific Time Series Encoder Pretraining framework via cross domain distillation. STEP introduces adaptive patching to handle extreme-length sequences and a statistics compensation scheme to accommodate diverse numerical scales. It further leverages cross-domain distillation to integrate knowledge from multiple foundation models into a unified encoder. By combining complementary representations across different domains, STEP learns general-purpose and transferable features tailored for scientific signals. Experiments on seven scientific time series tasks demonstrate that STEP provides both an effective structure and an effective pretraining paradigm, taking a STEP toward scientific time series representation learning.

Time series foundation models (TSFMs) require diverse, real-world datasets to adapt across varying domains and temporal frequencies. However, current large-scale datasets predominantly focus on low-frequency time series with sampling intervals, i.e., time resolution, in the range of seconds to years, hindering their ability to capture the nuances of high-frequency time series data. To address this limitation, we introduce a novel dataset, msData, that captures millisecond-resolution wireless and traffic conditions from an operational 5G wireless deployment, expanding the scope of TSFMs to incorporate high-frequency data for pre-training. Further, the dataset introduces a new domain, namely, wireless networks, thus complementing existing more general domains like energy and finance. The dataset also provides use cases for short-term forecasting, with prediction horizons spanning from 1 millisecond (1 step) to 96 milliseconds (96 steps). To demonstrate the utility of the dataset, we benchmark traditional machine learning models, transformer-based deep learning models, and TSFMs on forecasting tasks using representative subsets of the data, including a static mobility pattern within YouTube traffic class and a train mobility pattern within Web Browsing traffic class. Across these data distributions, we demonstrate that most TSFM model configurations perform poorly in both zero-shot and fine-tuned settings. Our work underscores the importance of incorporating high-frequency datasets during pre-training and forecasting to enhance architectures, fine-tuning strategies, generalization, and robustness of TSFMs in real-world applications. Code is available at this repository: https://github.com/khanalsubina/msData. Data is available on: Hugging Face.