Pith. sign in

REVIEW 19 cited by

Unified Training of Universal Time Series Forecasting Transformers

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.02592 v2 pith:N7DZCOL2 submitted 2024-02-04 cs.LG cs.AI

Gerald Woo , Chenghao Liu , Akshat Kumar , Caiming Xiong , Silvio Savarese , Doyen Sahoo This is my paper
classification cs.LGcs.AI
keywords seriestimeforecastingdatamodeluniversaladdressingchallenges
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Deep learning for time series forecasting has traditionally operated within a one-model-per-dataset framework, limiting its potential to leverage the game-changing impact of large pre-trained models. The concept of universal forecasting, emerging from pre-training on a vast collection of time series datasets, envisions a single Large Time Series Model capable of addressing diverse downstream forecasting tasks. However, constructing such a model poses unique challenges specific to time series data: i) cross-frequency learning, ii) accommodating an arbitrary number of variates for multivariate time series, and iii) addressing the varying distributional properties inherent in large-scale data. To address these challenges, we present novel enhancements to the conventional time series Transformer architecture, resulting in our proposed Masked Encoder-based Universal Time Series Forecasting Transformer (Moirai). Trained on our newly introduced Large-scale Open Time Series Archive (LOTSA) featuring over 27B observations across nine domains, Moirai achieves competitive or superior performance as a zero-shot forecaster when compared to full-shot models. Code, data, and model weights can be found at https://github.com/SalesforceAIResearch/uni2ts.

Discussion (0). Sign in to comment.

Forward citations

Cited by 19 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 32 citations worldwide. Full citation record

  1. TelecomTS: A Multi-Modal Observability Dataset for Time Series and Language Analysis

    cs.AI 2025-10 conditional novelty 7.0 of 10

    TelecomTS is a new observability dataset from 5G networks that preserves absolute scale and supports multi-modal tasks, showing that current time series and language models struggle with abrupt noisy dynamics.

  2. Byte Pair Encoding for Efficient Time Series Forecasting

    cs.LG 2025-05 conditional novelty 7.0 of 10

    A byte-pair-encoding tokenizer that converts repeated temporal motifs into single tokens improves zero-shot forecasting accuracy and speed over sample-wise and patch-based methods.

  3. A Benchmark for Electrical Load Forecasting Across Grid Levels: Time-Series Transformers Outperform Established Methods

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Transformers—especially a standard encoder-decoder—yield the lowest hourly load forecast errors across TSO, low-voltage feeder, and client-level datasets, with 6.6–10.7% error reduction over the best non-Transformer baseline.

  4. Trend strength predicts when generative foundation models win: a power-controlled benchmark, a mechanism, and an actionable selection rule

    stat.AP 2026-07 conditional novelty 6.0 of 10

    Zero-shot Chronos wins time-series benchmarks by under-extrapolating trend, and trend strength computed before forecasting predicts when it will beat classical models.

  5. WindFM: An Open-Source Foundation Model for Zero-Shot Wind Power Forecasting

    cs.LG 2025-09 conditional novelty 6.0 of 10

    A wind-specific foundation model, WindFM, uses hierarchical tokenization and autoregressive pre-training on the NREL WIND Toolkit to achieve state-of-the-art zero-shot wind power forecasts.

  6. STARE: Predicting Decision Making Based on Spatio-Temporal Eye Movements

    cs.NE 2025-08 unverdicted novelty 6.0 of 10

    A deep learning architecture that turns raw eye-tracking coordinates into spatial tokens for a time-series foundation model to predict consumer choices.

  7. Time Series Foundation Models for Multivariate Financial Time Series Forecasting

    q-fin.GN 2025-07 reject novelty 6.0 of 10

    Pretrained TTM shows large transfer and sample-efficiency gains in three financial forecasting tasks relative to training from scratch, but methodological flaws including possible look-ahead bias weaken the quantitati...

  8. Beyond Sensor Data: Foundation Models of Behavioral Data from Wearables Improve Health Predictions

    cs.LG 2025-06 conditional novelty 6.0 of 10

    A foundation model of wearable behavioral data outperforms simple baselines and complements a PPG sensor model across 57 health detection tasks.

  9. LightGTS: A Lightweight General Time Series Forecasting Model

    cs.LG 2025-06 conditional novelty 6.0 of 10

    A lightweight time series foundation model using period-aligned patches and parallel decoding reports zero-shot and full-shot accuracy on nine benchmarks comparable to much larger models.

  10. Modular Foundation Models for Time-Series Perception in Digital Twins

    cs.LG 2026-07 conditional novelty 5.0 of 10

    A gated bank of frozen self-supervised time-series encoders, aligned and aggregated by a Transformer, supports competitive multi-task perception for digital twins and hydro-generator virtual sensing.

  11. FLDmamba: Integrating Fourier and Laplace Transform Decomposition with Mamba for Enhanced Time Series Prediction

    cs.LG 2025-07 conditional novelty 5.0 of 10

    FLDmamba combines a learnable Fourier filter on Mamba's step size with a damped-sinusoid output layer and reports superior long-term forecasting accuracy on standard benchmarks.

  12. Towards Interpretable Time Series Foundation Models

    cs.CL 2025-07 conditional novelty 5.0 of 10

    After fine-tuning on 180 synthetic mean-reverting series annotated by a large multimodal model, small Qwen models can describe trend direction, noise intensity, and extremum location in natural language.

  13. Random Initialization Can't Catch Up: The Advantage of Language Model Transfer for Time Series Forecasting

    cs.CL 2025-06 conditional novelty 5.0 of 10

    Pretrained T5 language weights give a persistent validation-loss advantage over random initialization for low-data time series forecasting, and the advantage does not vanish within the training budget.

  14. Benchmarking Pre-Trained Time Series Models for Electricity Price Forecasting

    cs.LG 2025-06 conditional novelty 5.0 of 10

    No time series foundation model statistically outperforms the biseasonal MSTL model in most European day-ahead electricity price markets in 2024, though Chronos-Bolt and Time-MoE match traditional methods.

  15. Probabilistic Forecasting for Building Energy Systems using Time-Series Foundation Models

    cs.LG 2025-05 conditional novelty 5.0 of 10

    Fine-tuned time-series foundation models, especially Chronos with LoRA, outperform trained-from-scratch deep forecasters on multi-signal building energy forecasting with limited data.

  16. When can isotropy help adapt LLMs' next word prediction to numerical domains?

    cs.CL 2025-05 reject novelty 5.0 of 10

    Using a log-linear model and Jacobian analysis, the paper claims isotropy in LLM hidden embeddings stabilizes the softmax partition function and improves time-series forecasting, with illustrative experiments on five ...

  17. QuantFlow: A Federated Mamba-Based Post-Transformer Foundation Model for Time-Series Forecasting

    cs.LG 2026-07 reject novelty 4.5 of 10

    QuantFlow combines inverted embeddings, bidirectional Mamba decoders, quantile regression, TSMixup, and three-round FedAvg to report competitive one-step MSE on ETTm1/Weather and retain accuracy across 20 non-IID clients.

  18. Contextual Deconvolution for Variance-Stable Demand Sensing: Kernel-Modulated Operators in Promotional Retail

    cs.LG 2026-07 conditional novelty 4.0 of 10

    A smooth-baseline-plus-sparse-shock decomposition lowers forecast variance and safety stock but increases stockout costs, reducing total inventory cost only when holding costs exceed ~20% of stockout costs.

  19. On Identifying Why and When Foundation Models Perform Well on Time-Series Forecasting Using Automated Explanations and Rating

    cs.LG 2025-08 conditional novelty 4.0 of 10

    On four public datasets, Gradient Boosting with hand-built features beat Chronos, Llama, and ARIMA on most accuracy metrics, while Chronos only led on financial sMAPE.

Pith tools