Electronic health records (EHRs) are a rich source of real-world clinical data, but turning them into valid analyses remains slow, brittle, and expert-intensive. Although recent AI agents can answer medical questions and use tools, automating full EHR workflows remains difficult because small planning or execution erro...
Yinghao Zhu, Zi-Xiang Wang, Yi-Fan Qi et al.· npj Digital Medicine· 0 citations
StreamMind is introduced, a two-tier architecture that assigns latency-critical interaction and proactive monitoring to independently scheduled frontend workers, while backend workers asynchronously construct persistent multimodal memory and perform historical recall and external search.
Xichen Zhang, Guankai Li, Yinghao Zhu et al.· 1 citation
A diagnostic framework for AI agent behavior: layer attribution is proposed, which clarifies three consequences: surrogate validity is a model-task-layer relation, human-AI divergence provides diagnostic evidence, and governance requires source attribution before intervention.
