Evaluating Time-Series Foundation Models and Multimodal Dietary Context for CGM Forecasting
Abstract
Continuous glucose monitoring (CGM) provides high-frequency measurements of glucose dynamics and enables short-term glucose forecasting for diabetes management. Although time-series foundation models have shown strong general forecasting ability, their effectiveness for CGM prediction and the added value of multimodal dietary context remain unclear. We conduct a comprehensive empirical study using eight public CGM datasets spanning Type 1 diabetes, Type 2 diabetes, and non-diabetes populations. Under a unified protocol across multiple context lengths and prediction horizons, zero-shot foundation models did not consistently outperform strong task-specific baselines such as Elastic Net and PatchTST. In contrast, lightweight fine-tuning substantially improved forecasting performance. For example, fine-tuned Chronos-Bolt reduced RMSE by 6.5%-18.4% in the T1D cohort and by 8.6%-18.2% in the non-diabetes/T2D cohort, with comparable improvements in both in-distribution and out-of-distribution test settings. We further evaluate multimodal dietary context using CGMacros, which provides temporally aligned CGM signals, food images, and macronutrient records. A residual-based fusion framework reduced overall RMSE by approximately 3% and postprandial RMSE by approximately 15% relative to the CGM-only baseline. Moreover, Chronos-based CGM representations were more strongly correlated with observed postprandial glucose increments than representations from LSTM and CatBoost, even after those models incorporated additional dietary modalities, suggesting that pretrained temporal representations better preserve meal-induced excursion patterns. These findings show that foundation models require CGM-specific adaptation for reliable forecasting and that dietary context provides clinically meaningful signals beyond CGM alone, especially during postprandial periods.
Related
- SGA: Uncertainty Quantification for Multi-Step Forecasting in Time Series Foundation Models
- RACE: Residual-Aware Test-Time Adaptation for Neighbor-Rich Time-Series Foundation Model Forecasting
- Modular Bayesian Foundation Models: Bayesian time series encoders for OOD detection, routing and forecasting
- Event-Aware Spatiotemporal Precipitation Forecasting with Geographic Context and Physics-Guided Regularization
- Chronos-2: A foundation model for multivariate time series forecasting
BibTeX
@misc{zhang2026evaluating,
title = {{Evaluating Time-Series Foundation Models and Multimodal Dietary Context for CGM Forecasting}},
author = {Bowen Zhang and Hsiu-Wen Cheng and Hongyu Yang and Evie L. Shen and Joleen Vansomphone and Yuna Li and Kerry Zhou and Zitian Qu and Suning Zhao and Xiangning Deng and Hua Zhou and Jin J. Zhou},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.11872},
eprint = {2609.11872},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.11872},
}
RIS
TY - UNPB
TI - Evaluating Time-Series Foundation Models and Multimodal Dietary Context for CGM Forecasting
AU - Bowen Zhang
AU - Hsiu-Wen Cheng
AU - Hongyu Yang
AU - Evie L. Shen
AU - Joleen Vansomphone
AU - Yuna Li
AU - Kerry Zhou
AU - Zitian Qu
AU - Suning Zhao
AU - Xiangning Deng
AU - Hua Zhou
AU - Jin J. Zhou
PY - 2026
DA - 2026/09/10
PB - arXiv
DO - 10.48550/arxiv.2609.11872
UR - https://arxiv.org/abs/2609.11872
AB - Continuous glucose monitoring (CGM) provides high-frequency measurements of glucose dynamics and enables short-term glucose forecasting for diabetes management. Although time-series foundation models have shown strong general forecasting ability, their effectiveness for CGM prediction and the added value of multimodal dietary context remain unclear. We conduct a comprehensive empirical study using eight public CGM datasets spanning Type 1 diabetes, Type 2 diabetes, and non-diabetes populations. Under a unified protocol across multiple context lengths and prediction horizons, zero-shot foundation models did not consistently outperform strong task-specific baselines such as Elastic Net and PatchTST. In contrast, lightweight fine-tuning substantially improved forecasting performance. For example, fine-tuned Chronos-Bolt reduced RMSE by 6.5%-18.4% in the T1D cohort and by 8.6%-18.2% in the non-diabetes/T2D cohort, with comparable improvements in both in-distribution and out-of-distribution test settings. We further evaluate multimodal dietary context using CGMacros, which provides temporally aligned CGM signals, food images, and macronutrient records. A residual-based fusion framework reduced overall RMSE by approximately 3% and postprandial RMSE by approximately 15% relative to the CGM-only baseline. Moreover, Chronos-based CGM representations were more strongly correlated with observed postprandial glucose increments than representations from LSTM and CatBoost, even after those models incorporated additional dietary modalities, suggesting that pretrained temporal representations better preserve meal-induced excursion patterns. These findings show that foundation models require CGM-specific adaptation for reliable forecasting and that dietary context provides clinically meaningful signals beyond CGM alone, especially during postprandial periods.
ER -