{"slug": "discrete-neural-interlingua-for-ontology-agnostic-ehr-modeling", "title": "Discrete Neural Interlingua for Ontology-Agnostic EHR Modeling", "summary": "A new arXiv paper (2610.27047v1) introduces MEDIATOR, a discrete ontology-agnostic tokenizer that maps any clinical concept expressed as text into a shared vector-quantized codebook, replacing fixed-vocabulary EHR tokenization with open-vocabulary tokenization. Used as the token interface of an EHR Transformer, MEDIATOR improves mean AUPRC over the strongest prior tokenizer by 4.0 points (8% relative) on 20 MIMIC-III tasks and by 8.6 points (29% relative) in zero-shot transfer to 14 eICU tasks, where it uses all events; the authors note a closed-vocabulary tokenizer trained on MIMIC-III can use only 13% of eICU events. Under identical inputs MEDIATOR still leads by 1.0 and 1.9 points on operational and subcohort-mortality tasks, and the authors report same-encoder controls showing the text encoder drives in-domain accuracy while MEDIATOR's learned objectives drive transfer robustness and entity-linking quality.", "body_md": "arXiv:2610.27047v1 Announce Type: new \nAbstract: EHR foundation models promise scalable clinical prediction across health systems, yet their deployment is constrained by a tokenization bottleneck. The same clinical concept can appear as different ontology codes, local identifiers, synonyms, or free-text descriptions, which standard EHR tokenizers treat as unrelated symbols. A model trained on MIMIC-III with a closed-vocabulary tokenizer can use only 13% of eICU events. We introduce MEDIATOR, a discrete ontology-agnostic tokenizer that maps any clinical concept expressed as text into a shared vector-quantized codebook, turning EHR harmonization into open-vocabulary tokenization without manual mappings or a fixed vocabulary. Trained with synonym supervision and guided by enriched concept descriptions, code co-occurrence, and relation graphs, MEDIATOR collapses equivalent expressions while organizing distinct concepts in a medically meaningful discrete space. Used as the token interface of an EHR Transformer, MEDIATOR improves mean AUPRC over the strongest prior tokenizer by 4.0 points (8% relative) on 20 MIMIC-III tasks and by 8.6 points (29% relative) in zero-shot transfer to 14 eICU tasks, where it uses all events. The transfer gain combines coverage and representation, as MEDIATOR remains the best tokenizer under identical inputs (+1.0 and +1.9 points on operational and subcohort-mortality tasks). Same-encoder controls show that the text encoder drives in-domain accuracy, while the learned objectives of MEDIATOR drive transfer robustness and entity-linking quality. MEDIATOR also enables transferable attribution across hospitals.", "url": "https://wpnews.pro/news/discrete-neural-interlingua-for-ontology-agnostic-ehr-modeling", "canonical_source": "https://arxiv.org/abs/2610.27047", "published_at": "2026-10-21 04:00:00+00:00", "updated_at": "2026-10-21 04:18:20.014190+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "ai-research", "natural-language-processing"], "entities": ["MEDIATOR", "MIMIC-III", "eICU", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/discrete-neural-interlingua-for-ontology-agnostic-ehr-modeling", "markdown": "https://wpnews.pro/news/discrete-neural-interlingua-for-ontology-agnostic-ehr-modeling.md", "text": "https://wpnews.pro/news/discrete-neural-interlingua-for-ontology-agnostic-ehr-modeling.txt", "jsonld": "https://wpnews.pro/news/discrete-neural-interlingua-for-ontology-agnostic-ehr-modeling.jsonld"}}