One-Embedding-Fits-All: Efficient Zero-Shot Time Series Forecasting by a Model Zoo
This work addresses the challenge of leveraging complementary TSFMs for improved zero-shot forecasting in time series analysis, representing an incremental advancement by integrating existing models rather than introducing a fundamentally new method.
The paper tackles the problem of no single time series foundation model (TSFM) excelling universally in zero-shot forecasting by proposing ZooCast, which assembles a model zoo to dynamically select optimal models for different tasks, achieving strong performance on the GIFT-Eval benchmark while maintaining efficiency comparable to a single TSFM.
The proliferation of Time Series Foundation Models (TSFMs) has significantly advanced zero-shot forecasting, enabling predictions for unseen time series without task-specific fine-tuning. Extensive research has confirmed that no single TSFM excels universally, as different models exhibit preferences for distinct temporal patterns. This diversity suggests an opportunity: how to take advantage of the complementary abilities of TSFMs. To this end, we propose ZooCast, which characterizes each model's distinct forecasting strengths. ZooCast can intelligently assemble current TSFMs into a model zoo that dynamically selects optimal models for different forecasting tasks. Our key innovation lies in the One-Embedding-Fits-All paradigm that constructs a unified representation space where each model in the zoo is represented by a single embedding, enabling efficient similarity matching for all tasks. Experiments demonstrate ZooCast's strong performance on the GIFT-Eval zero-shot forecasting benchmark while maintaining the efficiency of a single TSFM. In real-world scenarios with sequential model releases, the framework seamlessly adds new models for progressive accuracy gains with negligible overhead.