T3Time: Tri-Modal Time Series Forecasting via Adaptive Multi-Head Alignment and Residual Fusion

TL;DR

T3Time introduces a tri-modal framework for multivariate time series forecasting that adaptively combines temporal, spectral, and prompt features, outperforming existing models especially in low-data scenarios.

Contribution

The paper proposes T3Time, a novel multi-branch model with adaptive multi-head alignment and residual fusion, enhancing horizon-specific and intervariable relationship modeling in time series forecasting.

Findings

Achieves 3.28% lower MSE and 2.29% lower MAE than state-of-the-art.

Demonstrates strong few-shot learning performance with significant error reductions.

Outperforms baselines across multiple benchmark datasets.

Abstract

Multivariate time series forecasting (MTSF) seeks to model temporal dynamics among variables to predict future trends. Transformer-based models and large language models (LLMs) have shown promise due to their ability to capture long-range dependencies and patterns. However, current methods often rely on rigid inductive biases, ignore intervariable interactions, or apply static fusion strategies that limit adaptability across forecast horizons. These limitations create bottlenecks in capturing nuanced, horizon-specific relationships in time-series data. To solve this problem, we propose T3Time, a novel trimodal framework consisting of time, spectral, and prompt branches, where the dedicated frequency encoding branch captures the periodic structures along with a gating mechanism that learns prioritization between temporal and spectral features based on the prediction horizon. We also proposed a mechanism which adaptively aggregates multiple cross-modal alignment heads by dynamically weighting the importance of each head based on the features. Extensive experiments on benchmark datasets demonstrate that our model consistently outperforms state-of-the-art baselines, achieving an average reduction of 3.28% in MSE and 2.29% in MAE. Furthermore, it shows strong generalization in few-shot learning settings: with 5% training data, we see a reduction in MSE and MAE by 4.13% and 1.91%, respectively; and with 10% data, by 3.62% and 1.98% on average. Code - https://github.com/monaf-chowdhury/T3Time/