CGM-JEPA: Learning Consistent Continuous Glucose Monitor Representations via Predictive Self-Supervised Pretraining
Summary
Introduces CGM-JEPA, a self-supervised pretraining framework for continuous glucose monitor data that improves cross-modal and cross-cohort performance through masked latent prediction and distributional objectives.
View Cached Full Text
Cached at: 05/11/26, 02:53 PM
Paper page - CGM-JEPA: Learning Consistent Continuous Glucose Monitor Representations via Predictive Self-Supervised Pretraining
Source: https://huggingface.co/papers/2605.00933
Abstract
A self-supervised pretraining framework for continuous glucose monitoring data achieves superior cross-modal and cross-cohort performance by predicting masked latent representations and incorporating cross-view distributional objectives.
Continuous Glucose Monitoring (CGM) can detect early metabolic subphenotypes (insulin resistance, IR; β-cell dysfunction), but population-scale deployment faces two coupled problems. First, the same physiological state appears through multiple views (CGM time series, venous OGTT, Glucodensity summaries), so single-view representations fail to transfer when deployment shifts the modality or setting. Second, baselines perform inconsistently across these shifts. Both problems point to one remedy: representations that abstract away from any single view to capture higher-level temporal and distributional structure. We propose CGM-JEPA, aself-supervised pretrainingframework which predictsmasked latent representationsrather than raw values, yielding abstraction that transfers across modalities. X-CGM-JEPA adds a masked Glucodensitycross-view objectivefor complementarydistributional information. We pretrain on sim389k unlabeled CGM readings from 228 subjects and evaluate on two clinical cohorts (N=27 and N=17 public-release subsets) across three regimes (cohort generalization,venous-to-CGM transfer, home CGM) under 20-iteration times 2-fold cross-validation. X-CGM-JEPA ranks first or second on AUROC for both endpoints across all three regimes while no baseline does, exceeding the strongest baseline by up to +6.5 pp incohort generalizationand +3.6 pp invenous-to-CGM transfer(paired Wilcoxon, p<0.001). Under modality shift, it matches mean AUROC while redistributing toward weaker subgroups (ethnicity AUROC gap shrinks 25-54%); on sparse in-domain venous data, the distributional view liftslabel-aware clustering(ARI+39%,NMI+40%). Code and weights: https://github.com/cruiseresearchgroup/CGM-JEPA
View arXiv pageView PDFGitHub1Add to collection
Get this paper in your agent:
hf papers read 2605\.00933
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2605.00933 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2605.00933 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2605.00933 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
CF-JEPA: Mask-free forward prediction with asymmetric encoder utilization for time-series representation learning
Proposes CF-JEPA, a mask-free self-supervised learning framework for time-series that uses multi-horizon forward prediction from random crops and exploits asymmetry between online and target encoders for improved performance on classification, forecasting, and anomaly detection.
STST-JEPA: Shallow-Target Spatio-Temporal Joint Embedding Prediction Architecture For EEG Self-Supervised Learning
Introduces STST-JEPA, a self-supervised transformer for EEG that predicts masked-token representations, pretrained on 47,703 sessions for brain age regression across ages 5–81.
@iScienceLuvr: Learning Sparse Latent Predictive Foundation Model for Multimodal Neuroimaging This paper introduces Neuro-JEPA, a foun…
This paper introduces Neuro-JEPA, a foundation model that uses a latent predictive objective and Mixture-of-Experts architecture to encode brain MRI scans across T1w, T2w, and FLAIR sequences, pretrained on a large dataset of 1.55 million scans.
HEPA: A Self-Supervised Horizon-Conditioned Event Predictive Architecture for Time Series
This paper introduces HEPA, a self-supervised architecture for predicting rare critical events in time series using a Joint-Embedding Predictive Architecture (JEPA) pretraining strategy. It demonstrates superior performance across multiple domains with significantly fewer labeled data and tuned parameters compared to leading models.
Predict and Reconstruct: Joint Objectives for Self-Supervised Language Representation Learning
The paper proposes a hybrid pre-training objective combining JEPA latent-space prediction with MLM reconstruction for language models, showing improved embedding uniformity and semantic-lexical balance.