Date
August 10, 2026
03
Presenters
In-person presenters
Contributors
Advances, challenges, and future directions — moving waveform AI from fluent prediction to conclusions grounded in explicit, replayable evidence.
Physiological waveforms, such as ECG, PPG, and EEG, provide high-fidelity views of underlying biological dynamics. Reliable interpretation depends on localized morphology, temporal structure, cross-channel consistency, and signal-quality awareness. Yet most current AI systems remain prediction-centric: they produce labels or fluent explanations without exposing the measurable evidence that supports intermediate claims and final decisions.
This tutorial develops a broader path toward verifiable physiological waveform reasoning. We connect advances in temporal modeling and time-series foundation models with data-driven and physics-informed approaches to dynamical system modeling—including graph neural ordinary differential equations and generative AI—and with deep learning methods tailored to health time series, such as physiological foundation models and multimodal learning. Building on these perspectives, we examine current signal, data, and model limitations and outline future directions in waveform–language alignment, agentic Plan–Act–Verify reasoning, and verification-centered evaluation. The goal is to move waveform AI from fluent prediction toward conclusions that are explicit, traceable, and replayable.
The design principles and evaluation interfaces generalize to other real-world time-series domains — wearables, industrial sensing, and IoT. No specialized clinical background is assumed.
@inproceedings{wang2026beyond,
title = {Beyond Prediction: Toward Verifiable Physiological Waveform
Reasoning: Advances, Challenges, and Future Directions},
author = {Wang, Xiaoda and Chang, Ching and Cao, Defu and Han, Kaiqiao
and Sun, Fang and Huang, Yue and Wang, Minxiao and Xu, Chang
and Luo, Xiao and Yan, Runze and Zhang, Xiangliang and Hu, Xiao
and Liu, Yan and Sun, Yizhou and Wang, Wei and Yang, Carl},
booktitle = {Proceedings of the 32nd ACM SIGKDD Conference on Knowledge
Discovery and Data Mining V.2 (KDD '26)},
year = {2026},
publisher = {ACM},
doi = {10.1145/3770855.3816466}
}
@inproceedings{wang2026position,
title = {Position: Beyond Prediction: Toward Verifiable Physiological
Waveform Reasoning with Foundation Models and Agentic LLMs},
author = {Wang, Xiaoda and Chang, Ching and Cao, Defu and Han, Kaiqiao
and Sun, Fang and Huang, Yue and Wang, Minxiao and Xu, Chang
and Luo, Xiao and Yan, Runze and Zhang, Xiangliang and Hu, Xiao
and Liu, Yan and Sun, Yizhou and Wang, Wei and Yang, Carl},
booktitle = {Proceedings of the 43rd International Conference on Machine
Learning (ICML)},
year = {2026}
}
@article{zhangdeep,
title = {Deep Learning for Health Time Series: Physiological Signals,
Clinical Records, and Epidemic Dynamics},
author = {Zhang, Lige and Huang, Ziwei and Wang, Xiaoda and Xu, Gelei
and Hu, Xiao and Yang, Carl and Jin, Wei},
year = {2026}
}