aiqbench analysis
This paper introduces VitalBench, a benchmark of 1,862 QA pairs and 90.2 hours of ECG/PPG recordings, and VitalAgent, a tool-augmented agent that pairs a longitudinal physiological memory with code-like computation over raw signals. The reported gain of over 25 points against prompt-based and agentic baselines is a benchmark result on the authors' own data, so it needs independent replication before it says much about real-world clinical use. For AI readers, the useful signal is the design pattern: giving an LLM agent tools for dynamic signal computation plus persistent memory, rather than relying on static summaries.
VitalAgent: A Tool-Augmented Agent for Reactive and Proactive Physiological Monitoring over Wearable Health Data
arXiv:2605.29483v3 Announce Type: replace Abstract: Wearable devices provide continuous access to physiological signals such as ECG and PPG, offering a path to answering users' questions about their own physiological state and to early detection of cardiac and physiological abnormalities. Yet existing mobile health (mHealth) systems are largely limited to task-specific prediction pipelines or reactive question answering over static summaries. They lack the temporal reasoning, persistent physiological context, and proactive monitoring required for long-term signal streams. To advance this setting, we introduce VitalBench, a longitudinal physiological monitoring benchmark comprising 1,862 QA pairs for reactive question answering and 90.2 hours of continuous ECG/PPG recordings for proactive monitoring, covering cardiac, physical activity, and stress-related tasks. Building on this resource, we propose VitalAgent, a tool-augmented agentic framework for ECG/PPG-based mHealth that supports both reactive question answering and proactive monitoring. VitalAgent is built on a longitudinal physiological memory and a tool-augmented reasoning interface that enables dynamic computation over raw signals. Experiments demonstrate that VitalAgent achieves over 25 points absolute improvement over prompt-based and agentic baselines in reactive evaluation and enables proactive alert monitoring over long-term physiological signals, pointing toward mHealth systems that can both answer questions about and continuously monitor a person's physiological state beyond the clinic.
Abstract:Wearable devices provide continuous access to physiological signals such as ECG and PPG, offering a path to answering users' questions about their own physiological state and to early detection of cardiac and physiological abnormalities. Yet existing mobile health (mHealth) systems are largely limited to task-specific prediction pipelines or reactive question answering over static summaries. They lack the temporal reasoning, persistent physiological context, and proactive monitoring required for long-term signal streams. To advance this setting, we introduce VitalBench, a longitudinal physiological monitoring benchmark comprising 1,862 QA pairs for reactive question answering and 90.2 hours of continuous ECG/PPG recordings for proactive monitoring, covering cardiac, physical activity, and stress-related tasks. Building on this resource, we propose VitalAgent, a tool-augmented agentic framework for ECG/PPG-based mHealth that supports both reactive question answering and proactive monitoring. VitalAgent is built on a longitudinal physiological memory and a tool-augmented reasoning interface that enables dynamic computation over raw signals. Experiments demonstrate that VitalAgent achieves over 25 points absolute improvement over prompt-based and agentic baselines in reactive evaluation and enables proactive alert monitoring over long-term physiological signals, pointing toward mHealth systems that can both answer questions about and continuously monitor a person's physiological state beyond the clinic.
Accepted at the GenAI for Health Workshop @ NeurIPS 2026 (GenAI4Health). v3: camera-ready version with revised proactive monitoring experiments. 25 pages, 3 figures, 13 tables
Artificial Intelligence (cs.AI)
arXiv:2605.29483 [cs.AI]
(or arXiv:2605.29483v3 [cs.AI] for this version)
Full text under the publisher's open licence · 233 words