Fine-grained List-wise Alignment for Generative Medication Recommendation

2025-05-26

AI Summary

Existing medical recommendation systems predominantly employ pointwise prediction, neglecting drug synergies and drug–drug interactions (DDIs), thereby compromising recommendation accuracy and safety in comorbid patient scenarios. To address this, we propose FLAME, a fine-grained listwise drug recommendation framework that models prescription generation as a sequential decision process of incremental drug addition and removal. FLAME introduces stepwise Groupwise Relative Policy Optimization (GRPO) coupled with potential-function-based reward shaping to achieve drug-level fine-grained alignment. It further enhances patient representation by jointly incorporating structured clinical knowledge and drug synergy information. Evaluated on multiple benchmark datasets, FLAME achieves state-of-the-art performance—significantly improving both recommendation accuracy and DDI safety—while enabling controllable trade-offs between safety and accuracy and demonstrating strong generalization across diverse clinical settings.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Learning Preferences or Rankings
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Accurate and safe medication recommendations are critical for effective clinical decision-making, especially in multimorbidity cases. However, existing systems rely on point-wise prediction paradigms that overlook synergistic drug effects and potential adverse drug-drug interactions (DDIs). We propose FLAME, a fine-grained list-wise alignment framework for large language models (LLMs), enabling drug-by-drug generation of drug lists. FLAME formulates recommendation as a sequential decision process, where each step adds or removes a single drug. To provide fine-grained learning signals, we devise step-wise Group Relative Policy Optimization (GRPO) with potential-based reward shaping, which explicitly models DDIs and optimizes the contribution of each drug to the overall prescription. Furthermore, FLAME enhances patient modeling by integrating structured clinical knowledge and collaborative information into the representation space of LLMs. Experiments on benchmark datasets demonstrate that FLAME achieves state-of-the-art performance, delivering superior accuracy, controllable safety-accuracy trade-offs, and strong generalization across diverse clinical scenarios. Our code is available at https://github.com/cxfann/Flame.

Problem

• Overcoming point-wise prediction limitations in medication recommendation systems
• Addressing synergistic drug effects and adverse drug-drug interactions (DDIs)
• Enhancing patient modeling with structured clinical knowledge and LLMs

Innovation

• Fine-grained list-wise alignment for LLMs
• Step-wise Group Relative Policy Optimization
• Integrates clinical knowledge into LLMs