Zero-shot Multivariate Time Series Forecasting Using Tabular Prior Fitted Networks

Mayuka Jayawardhana; Nihal Sharma; Kazem Meidani; Bayan Bruss; Tom Goldstein; Doron Bergman

cs.LGApr 9, 2026

Zero-shot Multivariate Time Series Forecasting Using Tabular Prior Fitted Networks

Authors: Mayuka Jayawardhana, Nihal Sharma, Kazem Meidani, Bayan Bruss, Tom Goldstein, Doron Bergman

Organizations: University of Maryland · Capital One

Abstract

Tabular foundation models, particularly Prior-data Fitted Networks like TabPFN have emerged as the leading contender in a myriad of tasks ranging from data imputation to label prediction on the tabular data format surpassing the historical successes of tree-based models. This has led to investigations on their applicability to forecasting time series data which can be formulated as a tabular problem. While recent work to this end has displayed positive results, most works have limited their treatment of multivariate time series problems to several independent univariate time series forecasting subproblems, thus ignoring any inter-channel interactions. Overcoming this limitation, we introduce a generally applicable framework for multivariate time series forecasting using tabular foundation models. We achieve this by recasting the multivariate time series forecasting problem as a series of scalar regression problems which can then be solved zero-shot by any tabular foundation model with regression capabilities. We present results of our method using the TabPFN-TS backbone and compare performance with the current state of the art tabular methods.

Figures & tables

Figure 1: Reformulating multivariate time series forecasting as a scalar regression problem : A pictorial and tabular representation of a Lorenz system (a time series with three covariates $x,y,z$ introduced in Lorenz (1963) ) is presented on the left parts of the figure. The time-indexed table is “rolled out” by introducing a channel indicator column (as seen in the table on the right). We note that this transformation expands the size of the table by a factor equal to the number of covariates.

Figure 2: Performance of TabPFN-TS compared to ours on multivariate datasets in the gift-eval benchmark our approach lowers MASE on 60% of the datasets. We refer the reader to Table 1 in the appendix for detailed results.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Dataset	MASE		WQL
Dataset	TabPFN-TS	TabPFN-TS-MV (ours)	TabPFN-TS	TabPFN-TS-MV (ours)
bizitobs_application/10S/long	3.3314	3.1884	0.0483	0.0490
bizitobs_application/10S/medium	2.8469	2.5012	0.0395	0.0415
bizitobs_application/10S/short	1.7608	1.4309	0.0178	0.0171
bizitobs_l2c/5T/medium	1.2231	1.0165	0.5009	0.4470
bizitobs_l2c/5T/short	0.4127	0.3548	0.1046	0.0973

Appendix

Table 1: Detailed performance comparison with TabPFN-TS. We evaluate our approach using the multivariate datsets in the Gift-Eval time series benchmark Aksu et al. (2024) . Following standard practice we evaluate point forecast accuracy using the Mean Absolute Scaled Error (MASE) and probabilistic forecast accuracy via the Weighted Quantile Loss (WQL). Lower is better for both metrics. Standard deviations are omitted, as variance across random seeds for TabPFN was negligible.

	MASE
	Chronos 2	TempoPFN	TimePFN	TabPFN-TS	TabPFN-TS-MV (ours)
Average	6.8068	1.2838	2.8429	1.2386	1.2255
Average (w/o jena)	1.0355	1.32823	3.00798	1.2718	1.2370
Average Rank	1.6452	2.8387	4.871	2.9677	2.6774

Appendix

Table 2: MASE comparison with SOTA Models (Lower is better): Average (w/o jena) denotes the average MASE across the multivariate datasets in the Gift-Eval Benchmark excluding the Jena Weather dataset Max Planck Institute for Biogeochemistry (2017) . We denote this metric due the poor performance of Chronos-2 specifically for this dataset.

	WQL
	Chronos 2	TempoPFN	TimePFN	TabPFN-TS	TabPFN-TS-MV (ours)
Average	0.2577	0.1511	0.3354	0.1541	0.1586
Average (w/o jena)	0.1392	0.1648	0.3659	0.1687	0.1721
Average Rank	1.5484	2.6129	4.871	2.9032	3.0645

Appendix

Table 3: WQL comparison with SOTA Models (Lower is better): Average (w/o jena) denotes the average WQL across the multivariate datasets in the Gift-Eval Benchmark excluding the Jena Weather dataset Max Planck Institute for Biogeochemistry (2017) . We denote this metric due the poor performance of Chronos-2 specifically for this dataset.

Figure 3: Performance of TabPFN-TS-MV compared to other SOTA time series approaches: The y-axis is log scale, Tabpfn-TS-MV (ours) and Chronos 2 supports channel dependent (CD) multivariate predictions while TabPFN-TS and TempoPFN only support univariate (channel independent) predictions.

Figure 5: Performance comparison between different strategies for mitigating cross-channel distributional shifts. The y-axis is in log scale. We observe that while the standardizing approach greatly improves performance, first order differencing degrades performance. We incorporate the standardization approach into our framework.

Multivariate Time Series Forecasting

Multivariate time series forecasting predicts future values across related variables by learning their temporal and cross-variable dependencies.

See all 299 papers

Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitives
Oct 8, 2026Haoran Zhang, Haixuan Liu, Xingjian Su +5Multivariate Time Series Forecasting Time-Series Foundation Models
School of Software, BNRist, Tsinghua University · ByteDance, Beijing, China
DualWeaver: Synergistic Feature Weaving Surrogates for Multivariate Forecasting with Univariate Time Series Foundation Models
Feb 25, 2026Jinpeng Li, Zhongyi Pei, Huaze Xue +3Multivariate Time Series Forecasting Time Series Forecasting
School of Software, BNRist, Tsinghua University · Tencent
Foundations without Fundamentals: Zero-Shot Blind Spots in Time Series FMs
Oct 1, 2026Nafiseh Ghoroghchian, Haipeng Zhang, Shuyi Han +2Multivariate Time Series Forecasting Zero-Shot Time Series Forecasting
Layer6 AI

Oct 8, 2026J/K move · Enter open · S save

Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitives

Haoran Zhang, Haixuan Liu, Xingjian Su, Yong Liu+4

School of Software, BNRist, Tsinghua University · ByteDance, Beijing, China

Multivariate Time Series Forecasting Time-Series Foundation Models Zero-Shot Time Series Forecasting Time Series Transformers Synthetic Data Pretraining

Open paper AskPDF arXiv

We introduce Timer-M1, a pretrained multivariate time series foundation model that learns with primitives for zero-shot forecasting. Across domains, time series share elementary temporal and relational patterns, termed primitives, yet differ in how these primitives manifest and evolve across different contexts. Despite progress in zero-shot and task-general forecasting, existing foundation models may still struggle to generalize to complex real-world scenarios. To this end, we develop a primitive-based data synthesis and pretraining pipeline. The synthesis pipeline generates series with temporal primitives shared across domains and then assembles real and generated series into multivariate samples using relational primitives. Afterwards, samples are organized into episodes by assigning distinct channel roles as target variates, past-only covariates, and known-future covariates, ensuring that the model is optimized on predictable variates using available exogenous information. Technically, Timer-M1 further adapts gated two-dimensional Transformer blocks that dynamically allocate cross-variate attention across layers. Across three large-scale forecasting benchmarks, Timer-M1 ranks first on both FEV and TIME and second on GIFT-Eval among most recent time series foundation models. These results support effective primitive-based pretraining as a route to robust general forecasting technique across domains and task settings.

Abstract

Multivariate Time Series Forecasting

Explore similar work

Timer-M1: A Multivariate Time Series Foundation Model via Learning Primitives

DualWeaver: Synergistic Feature Weaving Surrogates for Multivariate Forecasting with Univariate Time Series Foundation Models

Foundations without Fundamentals: Zero-Shot Blind Spots in Time Series FMs