TimeRouter: Efficient and Adaptive Routing of Time-Series Foundation Models

2026-06-09

AI Summary

This work addresses the inefficiency of existing time series foundation model (TSFM) pools, which lack adaptive expert selection mechanisms and incur high inference overhead due to reliance on large language model (LLM) controllers. To overcome this limitation, we propose TimeRouter—the first framework enabling LLM-free, adaptive routing among pretrained TSFMs. TimeRouter employs a lightweight discriminative routing head, selective gating, and an ensemble fallback mechanism to dynamically dispatch inputs to the most suitable models in the pool. This approach significantly enhances system modularity and inference efficiency, achieving state-of-the-art performance on the GIFT-EVAL benchmark with an LB MASE of 0.6765. Furthermore, our experiments demonstrate that both the composition of the model pool and the design of the gating mechanism critically influence routing effectiveness.

Abstract

Time-series foundation models (TSFMs) are increasingly explored as predictive experts within emerging agentic time-series systems. However, TSFMs exhibit heterogeneous inductive biases, and no single model consistently dominates across forecasting regimes, making expert selection a critical challenge. Existing systems often delegate this decision to LLM-based controllers, incurring substantial inference overhead. We present TimeRouter, an efficient routing framework that leverages empirical complementarity across a pool of pretrained TSFMs through lightweight discriminative routing, selective gating, and ensemble fallback. Concretely, TimeRouter combines a learned routing head, a selective gate, and an ensemble fallback, enabling adaptive expert selection without invoking an LLM at inference time. TimeRouter achieves state-of-the-art performance on the GIFT-EVAL leaderboard, with an LB MASE of 0.6765. Beyond benchmark performance, our ablation studies provide empirical insights into TSFM routing design, highlighting the importance of pool composition and selective gating. Taken together, these results position TimeRouter as a modular and lightweight routing layer for future agentic time-series systems built upon foundation-model pools. Our code is available at https://github.com/UConn-DSIS/TimeRouter.

Problem

• time-series foundation models
• expert selection
• model routing
• forecasting
• inductive bias

Innovation

• Time-Series Foundation Models
• Model Routing
• Selective Gating
• Ensemble Fallback
• Efficient Inference