Unicorn: Scaling High-Dimensional Time Series Forecasting via Universal Correlation Modeling

2026-05-26

AI Summary

Existing methods for high-dimensional time series forecasting face a generalization bottleneck in balancing channel independence and cross-channel dependency modeling. This work proposes UniCorN, a framework that leverages multi-dataset pretraining to construct a shared latent space and introduces a channel-identity-disentangled latent prototype codebook to learn transferable, identity-agnostic cross-channel interaction patterns. By decoupling channel-specific characteristics from the learned representations, UniCorN overcomes dimensional constraints and enables universal correlation modeling. The approach significantly outperforms current state-of-the-art methods across multiple benchmarks, demonstrating particularly strong performance in few-shot transfer scenarios. These results establish UniCorN as a promising pathway toward foundational models for multivariate time series analysis.

Abstract

Modern time series architectures face a fundamental trade-off: channel-independent models scale well with increasing data volume but ignore critical inter-channel dependencies, while channel-dependent models are expressive but remain ``dimension-bounded'', struggling to generalize across heterogeneous datasets.To bridge this gap, we introduce Unicorn (Universal Correlation Network), a framework for scalable, multi-dataset pretraining on high-dimensional time series. At the core of Unicorn is a latent prototype codebook that decouples correlation modeling from specific channel identities. By projecting heterogeneous channels into a shared latent space, UniCorN learns identity-agnostic, reusable interaction patterns that transfer across domains with diverse dimensionalities and semantics. Extensive experiments show that Unicorn significantly outperforms state-of-the-art forecasting architectures, particularly in few-shot transfer scenarios, offering a scalable path toward multivariate time series foundation models.

Problem

• time series forecasting
• high-dimensional
• inter-channel dependencies
• scalability
• cross-dataset generalization

Innovation

• universal correlation modeling
• latent prototype codebook
• channel-agnostic representation
• multivariate time series forecasting
• foundation model