RMISC: A Large-scale Real-world Multivariate Corpus for Time Series Foundation Models

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Qian Sun, Yong-Ming Tian, Jia-Wei Huang, Cheng Feng, Shao-Qun Zhang
• Venue: arXiv: cs.AI
• Published: 2026-07-07
• Collected: 2026-07-08
• URL: http://arxiv.org/abs/2607.06504v1
• PDF: https://arxiv.org/pdf/2607.06504v1
• DOI: UNKNOWN

Plain-Language Brief

The paper presents RMISC, a large-scale, high-quality, open-access corpus of real-world multivariate time series data. The researchers pretrain four advanced Time Series Foundation Models (TSFMs) on univariate, synthetic multivariate, and real-world multivariate data. Experimental results show that incorporating real-world multivariate data improves the generalization performance for both univariate and multivariate TSFMs.

What the Document Says

• The paper introduces RMISC, a large-scale real-world multivariate time series corpus containing around 200 datasets and 142 billion time points across various domains.
• Four advanced Time Series Foundation Models (TSFMs) are pretrained on univariate, synthetic multivariate, and real-world multivariate data.
• The zero-shot generalization capabilities of these models are evaluated on standard in-distribution and out-of-distribution benchmarks.
• Experimental results show that using real-world multivariate data improves the generalization performance for both univariate and multivariate TSFMs.

Key Concepts Mentioned

• Time Series Foundation Models (TSFMs)
• Real-world Multivariate Time Series Data
• Corpus
• Pretraining
• Generalization Performance

Claims Made by the Document

• The RMISC corpus is a large-scale, high-quality, open-access real-world multivariate time series archive.
• Incorporating real-world multivariate data improves the generalization performance for both univariate and multivariate TSFMs.

Evidence Used

• The introduction and description of RMISC corpus in the paper.
• Experimental results presented in the paper showing improved generalization performance with real-world multivariate data.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Recent years have witnessed the emergence of multivariate modeling using time series foundation models (TSFMs), which achieve advanced zero-shot generalization. Modern multivariate TSFMs are predominantly pretrained on multivariate synthetic data, which is easier to scale but may fail to capture the complex temporal dynamics and cross-variable relationships present in real-world time series. This raises a key question: Whether and to what extent the leading TSFMs trained with the real-world corpus perform better than those trained with synthetic data? To answer this, we establish the RMISC corpus, a considerably large-scale, high-quality, openly accessible, real-world, and multivariate time series archive that contains around 200 datasets and 142 billion time points across diverse domains. Furthermore, we pretrain four advanced TSFMs on univariate, synthetic multivariate, and real-world multivariate data and evaluate their zero-shot generalization capabilities on standard in-distribution and out-of-distribution benchmarks. Experimental results show that incorporating real-world multivariate data predominantly improves the generalization performance for both univariate and multivariate TSFMs. These results provide a deeper understanding of how real-world multivariate data contributes to the development of stronger TSFMs.