Clinical evidence indicates that (1) disease prevalence differs across racial groups and (2) medication prescriptions for identical diagnoses and procedures may vary across racial groups. However, existing medication recommendation methods have not integrated these racial properties into the design of their encoder and predictor architectures. To address this gap, we propose RaceMED, which incorporates properties (1) and (2) through a dual-branch encoder composed of race-specific and race-general branches that capture idiosyncratic features unique to each racial group and general patterns shared across all patients, and a predictor based on race-aware attention that restricts cross-patient visit references to patients from the same racial group, preventing inappropriate medication transfer across racial groups. Extensive experiments demonstrate that RaceMED consistently outperforms ten state-of-the-art competitors, achieving up to 12.24% improvement in accuracy while reducing performance disparities across racial groups by up to 31.77%, thereby improving fairness. These findings demonstrate that explicitly modeling racial properties is essential for improving both accuracy and fairness, a dimension that has been largely underexplored in the medication recommendation domain.
Hojung Shin, Taeri Kim, Jebum Choi et al.· Proceedings of the 32nd ACM...· 0 citations
Hypergraphs effectively model higher-order groupwise relationships beyond pairwise interactions, while pretrained language models (PLMs) and large language models (LLMs) provide rich semantic understanding from textual attributes. However, research on combining language models with hypergraph learning remains limited due to the lack of public text-attributed hypergraph benchmarks. To address this limitation, we present TAHB (Text-Attributed Hypergraph Benchmark), the first public benchmark integrating hypergraph structures and raw textual attributes. TAHB contains 10 real-world datasets from four domains - e-commerce, academia, movies, and politics networks - enabling systematic evaluation of text-aware hypergraph representation learning. Experimental results show that TAHB preserves key structural properties of real-world hypergraphs and consistently reproduces performance tendencies observed in existing benchmarks. Furthermore, experiments under both LLM-as-Enhancer and LLM-as-Predictor settings demonstrate that LLM-enhanced textual semantics improve hypergraph learning performance, while structural and textual information jointly provide the best setting for LLM-based prediction. Our benchmark provides a foundation for future research at the intersection of hypergraph learning and language models.
D. Y. Kang, JungHyun Kim, Juhyun Jeon et al.· 0 citations