PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
September 24, 20250 citationsOpen Access

UniCast: A Unified Multimodal Prompting Framework for Time Series Forecasting

View Full Paper
SPS. H. ParkSHSoyeon Caren HanEHEduard Hovy

Key Points

  • UniCast significantly enhances forecasting performance by leveraging multimodal inputs, including text and vision.
  • Experiments show consistent improvement across time-series benchmarks, validating the effectiveness of the unified approach.
  • The use of soft prompt tuning allows for optimal integration of pretrained embeddings with minimal updates.
  • The findings underscore the importance of multimodal data in developing advanced time series forecasting models.

Abstract

Time series forecasting is a foundational task across domains, such as finance, healthcare, and environmental monitoring. While recent advances in Time Series Foundation Models (TSFMs) have demonstrated strong generalisation through large-scale pretraining, existing models operate predominantly in a unimodal setting, ignoring the rich multimodal context, such as visual and textual signals, that often accompanies time series data in real-world scenarios. This paper introduces a novel parameter-efficient multimodal framework, UniCast, that extends TSFMs to jointly leverage time series, vision, and text modalities for enhanced forecasting performance. Our method integrates modality-specific embeddings from pretrained Vision and Text Encoders with a frozen TSFM via soft prompt tuning, enabling efficient adaptation with minimal parameter updates. This design not only preserves the generalisation strength of the foundation model but also enables effective cross-modal interaction. Extensive experiments across diverse time-series forecasting benchmarks demonstrate that UniCast consistently and significantly outperforms all existing TSFM baselines. The findings highlight the critical role of multimodal context in advancing the next generation of general-purpose time series forecasters.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Park et al. (2025) studied this question.

synapsesocial.com/papers/68d6e1978b2b6861e4c4033fhttps://doi.org/10.48550/arxiv.2508.11954
Ask AI
Helpful
Bookmark
Share
View Full Paper

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Dual-Forecaster: A Multimodal Time Series Model Integrating Descriptive and Predictive Texts2025
  2. 2GPT4MTS: Prompt-based Large Language Model for Multimodal Time-series Forecasting2024 · 93 citations
  3. 3UniTS: A Unified Multi-Task Time Series Model2024 · 3 citations
  4. 4U-Cast: Learning Hierarchical Structures for High-Dimensional Time Series Forecasting2025 · 1 citations
  5. 5TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment2024 · 3 citations