Clin-JEPA: A Multi-Phase Co-Training Framework for Joint-Embedding Predictive Pretraining on EHR Patient Trajectories
arXiv:2605.10840v5 Announce Type: replace-cross Abstract: Joint-embedding predictive architectures (JEPA) learn representations by predicting in latent space, as in computer vision; retaining the action-conditioned predictor at inference turns them…
Read the full story at arXiv cs.AI ↗
ImpactNotable 31/100
Why it mattersRule-based estimate: event keywords (+4); trust 6/10.
RegionsGlobal
Published9 h ago (Fri, 18 Sep 2026 04:00:00 GMT)
RetrievedFri, 18 Sep 2026 08:00:48 GMT via rss
ClassifiedFri, 18 Sep 2026 08:01:04 GMT by heuristic
AuthorYixuan Yang, Mehak Arora, Ryan Zhang, Baraa Abed, Junseob Kim, Tilendra Choudhary, Md Hassanuzzaman, Kevin Zhu, Ayman…