Yangyang Fu

Xing Lu, Kathryn Hinkelman, Yangyang Fu et al.

Infrastructure in future smart and connected communities is envisioned as an aggregate of public services, including the energy, transportation and communication systems, all intertwined with each other. The intrinsic interdependency among these systems may exert underlying influence on both design and operation of the heterogeneous infrastructures. However, few prior studies have tapped into the interdependency among the three systems in order to quantify their potential impacts during standard operation. In response to this, this paper proposes an open source, flexible, integrated modeling framework suitable for designing coupled energy, transportation, and communication systems and for assessing the impact of their interdependencies. First, a novel multi-level, multi-layer, multi-agent approach is proposed to enable flexible modeling of the interconnected energy, transportation, and communication systems. Then, for the framework's proof-of-concept, preliminary component and system-level models for different systems are designed and implemented using Modelica, an equation-based object-oriented modeling language. Finally, three case studies of gradually increasing complexity are presented (energy, energy + transportation, energy + transportation + communication) to evaluate the interdependencies among the three systems. Quantitative analyses show that the deviation of the average velocity on the road can be 10.5\% and the deviation of the power draw from the grid can be 7\% with or without considering the transportation and communication system at the peak commute time, indicating the presence of notable interdependencies. The proposed modeling framework also has the potential to be further extended for various modeling purposes and use cases, such as dynamic modeling and optimization, resilience analysis, and integrated decision making in future connected communities.

Anran Li, Lingfei Qian, Mengmeng Du et al.

Large Language Models (LLMs) have demonstrated significant potential in medicine. To date, LLMs have been widely applied to tasks such as diagnostic assistance, medical question answering, and clinical information synthesis. However, a key open question remains: to what extent do LLMs memorize medical training data. In this study, we present the first comprehensive evaluation of memorization of LLMs in medicine, assessing its prevalence (how frequently it occurs), characteristics (what is memorized), volume (how much content is memorized), and potential downstream impacts (how memorization may affect medical applications). We systematically analyze common adaptation scenarios: (1) continued pretraining on medical corpora, (2) fine-tuning on standard medical benchmarks, and (3) fine-tuning on real-world clinical data, including over 13,000 unique inpatient records from Yale New Haven Health System. The results demonstrate that memorization is prevalent across all adaptation scenarios and significantly higher than reported in the general domain. Memorization affects both the development and adoption of LLMs in medicine and can be categorized into three types: beneficial (e.g., accurate recall of clinical guidelines and biomedical references), uninformative (e.g., repeated disclaimers or templated medical document language), and harmful (e.g., regeneration of dataset-specific or sensitive clinical content). Based on these findings, we offer practical recommendations to facilitate beneficial memorization that enhances domain-specific reasoning and factual accuracy, minimize uninformative memorization to promote deeper learning beyond surface-level patterns, and mitigate harmful memorization to prevent the leakage of sensitive or identifiable patient information.