Generative recommendation (GR) reformulates sequential recommendation as an autoregressive generation problem, where items are represented as discrete semantic IDs. However, learning effective item tokenization is critical yet remains challenging. Most existing methods optimize tokenization in a point-wise or heuristic manner, which often leads to Relational Structure Distortion. We identify two main sources of this distortion: (1) point-wise reconstruction losses that are insufficient to preserve item--item relations, and (2) collision-mitigation--driven code assignment that disrupts semantic neighborhoods under many-to-one mappings. This issue is especially harmful for recommendation, since the task fundamentally depends on relative item similarity. To address this issue, we propose ReSOT, a unified framework that Re-balances Semantic ID learning via Optimal Transport for GR. ReSOT consists of two complementary components: a structure-aware reconstruction module that aligns item–item relations between the original and the reconstructed embedding space, and an unbalanced optimal transport-based soft quantization module that achieves a better trade-off between collision mitigation and semantic neighborhood preservation. Extensive experiments on real-world datasets show that ReSOT provides a principled tokenization scheme that preserves relational structure while assigning codes in a collision-aware and semantics-consistent manner. Our code is available at https://github.com/grw-zju/ReSOT.
Renwu Geng, Yiming Xu, Fengxin Li et al.· Proceedings of the 32nd ACM...· 0 citations
Large Language Models (LLMs) have shown strong potential in medical applications such as question answering and clinical prediction. % Despite their growing adoption, fairness in LLMs for medicine remains underexplored, largely due to the mismatch between conventional fairness constraints and the clinically meaningful role of sensitive attributes. Existing approaches often enforce attribute-invariant constraints, leading to substantial performance degradation that is unacceptable in high-stakes healthcare settings. Moreover, fairness evaluation for medical LLMs is hindered by the lack of dedicated benchmarks. In this paper, we first rethink fairness in LLMs for medicine from a clinically grounded, utility-based perspective. Inspired by principles of health equity in medicine, we introduce universal fairness, a clinically grounded definition that reframes fairness as maximizing subgroup-aware diagnostic performance under attribute-conditioned health disparities. To achieve this objective in practice, we propose MAPPE, a training-free minimax prompt optimization framework. % MAPPE theoretically promotes universal fairness, while directly applicable to both closed-source and open-source LLMs. To systematically evaluate fairness, we construct FairMed, the first attribute-annotated benchmark for medical LLMs covering medical question answering and clinical prediction. % Experiments on both closed-source and open-source LLMs reveal demographic disparities, while MAPPE consistently improves worst-group and overall performance, outperforming existing fairness-oriented and prompt-based methods. The dataset and code are available at https://github.com/xiye7lai/FairMed.
Jiaming Zhang, Yuyuan Li, Xiaohua Feng et al.· Proceedings of the 32nd ACM...· 0 citations
