TRACE: Temporal Reasoning via Agentic Context Evolution for Streaming Electronic Health Records (EHRs)
Zhan Qu, Michael Färber · 2026
Read TRACE: Temporal Reasoning via Agentic Context Evolution for Streaming Electronic Health Records (EHRs) by Zhan Qu, Michael Färber (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-12833.txt.
Description
Thinking over long hospital records strains language-model systems that keep re-reading whole past notes or passing free-form notes between agents. Zhan Qu and Michael Färber present Vital Trace, a multi-agent system bound by fixed rules for calling future care risk in ICU care. Instead of open-ended text past notes, it keeps a compact patient-state memory and splits thinking among four agents: a router, a thinker, an auditor, and a steward.
A hand-picked global rule set gives body change rules. A live state record tracks heart, lung, kidney, metabolic, and swelling trouble over time. On MIMIC-IV and eICU, the shaped approach raised time-based fit, steady talk, tuning, and clarity over free-form multi-agent plain rivals. It also kept strong forecast scores across long ICU stays.
Catalog record
| CREATOR | Zhan Qu, Michael Färber |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-12833