Selected work

Representative Papers

TCMClinicalReason-Bench: Can Language Models Reason from Pathogenesis to Prescription over Real-World Clinical Cases?

Oct 03, 2026

This study addresses the lack of evidence grounding and the cascading propagation of errors in clinical reasoning for Traditional Chinese Medicine (TCM) within large language models. To this end, it constructs a full-pipeline evaluation benchmark derived from real-world, multicenter electronic health records. Methodologically, the work proposes an evidence-constrained scoring mechanism that decouples component-level quality from cross-module logical consistency to precisely localize reasoning bottlenecks, while integrating automated judge models with expert blind reviews to ensure evaluation reliability. The findings reveal that general-purpose models outperform domain-specific counterparts and identify significant deficiencies in prescription generation. These insights provide critical empirical foundations for enhancing the robustness of TCM-oriented AI reasoning systems.

0 citationsRead paper

TCMaster: Confidence-Aware Querying and Workload-Guided Physical Design for Multi-Source Traditional Chinese Medicine Knowledge Graphs

Sep 22, 2026

TCMaster通过集成多源数据、标注置信度和优化查询策略，解决了中医药知识图谱中查询可靠性和性能问题。

0 citationsRead paper

Medical Test-free Disease Detection Based on Big Data

Nov 26, 2025

To address the high cost and limited scalability of clinical laboratory testing—particularly for screening hundreds to thousands of diseases—this paper proposes CLDD, a Graph Neural Collaborative Learning model for disease detection. CLDD reformulates disease detection as an adaptive collaborative learning task, jointly modeling disease–disease associations and patient–patient similarities, thereby eliminating reliance on disease-specific diagnostic tests. The model integrates heterogeneous features—including patient–disease interactions and demographic attributes—from electronic health records (EHRs) and employs a graph neural network for collaborative representation learning. Additionally, it incorporates an interpretable ranking mechanism to support clinical decision-making. Evaluated on the MIMIC-IV dataset (61,191 patients, 2,000 diseases), CLDD achieves absolute improvements of 6.33% in recall and 7.63% in precision over state-of-the-art baselines. It further demonstrates strong capability in recovering masked diseases and provides clinically meaningful, interpretable predictions.

0 citationsRead paper

Recent publications

Latest Papers

TCMClinicalReason-Bench: Can Language Models Reason from Pathogenesis to Prescription over Real-World Clinical Cases?

Oct 03, 2026

This study addresses the lack of evidence grounding and the cascading propagation of errors in clinical reasoning for Traditional Chinese Medicine (TCM) within large language models. To this end, it constructs a full-pipeline evaluation benchmark derived from real-world, multicenter electronic health records. Methodologically, the work proposes an evidence-constrained scoring mechanism that decouples component-level quality from cross-module logical consistency to precisely localize reasoning bottlenecks, while integrating automated judge models with expert blind reviews to ensure evaluation reliability. The findings reveal that general-purpose models outperform domain-specific counterparts and identify significant deficiencies in prescription generation. These insights provide critical empirical foundations for enhancing the robustness of TCM-oriented AI reasoning systems.

0 citationsRead paper

TCMaster: Confidence-Aware Querying and Workload-Guided Physical Design for Multi-Source Traditional Chinese Medicine Knowledge Graphs

Sep 22, 2026

TCMaster通过集成多源数据、标注置信度和优化查询策略，解决了中医药知识图谱中查询可靠性和性能问题。

0 citationsRead paper

Medical Test-free Disease Detection Based on Big Data

Nov 26, 2025

To address the high cost and limited scalability of clinical laboratory testing—particularly for screening hundreds to thousands of diseases—this paper proposes CLDD, a Graph Neural Collaborative Learning model for disease detection. CLDD reformulates disease detection as an adaptive collaborative learning task, jointly modeling disease–disease associations and patient–patient similarities, thereby eliminating reliance on disease-specific diagnostic tests. The model integrates heterogeneous features—including patient–disease interactions and demographic attributes—from electronic health records (EHRs) and employs a graph neural network for collaborative representation learning. Additionally, it incorporates an interpretable ranking mechanism to support clinical decision-making. Evaluated on the MIMIC-IV dataset (61,191 patients, 2,000 diseases), CLDD achieves absolute improvements of 6.33% in recall and 7.63% in precision over state-of-the-art baselines. It further demonstrates strong capability in recovering masked diseases and provides clinically meaningful, interpretable predictions.

0 citationsRead paper