---
title: "RMISC: A Large-scale Real-world Multivariate Corpus for Time Series Foundation Models"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.06504v1"
permalink: "https://feed.astralor.com/items/itm_02c24e08ea1820dcc9636ee7"
published: "2026-07-07T17:04:44.000Z"
score: 78
channel: "ai"
tags: ["RMISC","Multivariate Time Series","Time Series Foundation Models","Real-world data","zero-shot generalization"]
---

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

## AI commentary

> This paper is a shot in the arm for the time series foundation model community — it directly addresses the debate over whether synthetic data is sufficient by leveraging massive real-world data. For algorithm engineers and research teams working on time series forecasting, the RMISC corpus itself is a treasure worth prioritizing in their experimental toolkit. More importantly, the results imply that the next wave of TSFM competition may shift from model architecture to data quality.

## At a glance

Recent years have seen multivariate time series foundation models (TSFMs) predominantly pretrained on synthetic data, which fails to capture real-world complexities. This paper introduces RMISC, a large-scale corpus with about 200 datasets and 142 billion time points across diverse domains. By pretraining four advanced TSFMs on univariate, synthetic multivariate, and real-world multivariate data, experiments show that real-world multivariate data significantly improves zero-shot generalization for both univariate and multivariate TSFMs, providing crucial resources and empirical evidence for stronger models.

## Key points

- Constructed RMISC, a large-scale real-world multivariate time series corpus with ~200 datasets and 142 billion time points across domains
- Pretrained four advanced TSFMs on univariate, synthetic multivariate, and real-world multivariate data respectively
- Experiments show real-world multivariate data improves zero-shot generalization for both univariate and multivariate TSFMs
- Reveals the critical role of real data in capturing complex temporal dynamics and cross-variable relationships

## Full text

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.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_02c24e08ea1820dcc9636ee7