MediRec: Explainable LLM Framework for Chinese Medication Recommendation
Researchers have created a new system named MediRec, which employs large language models (LLMs) to suggest medications based on Chinese electronic health records (EHRs). This new approach addresses a significant gap in using LLMs in China’s clinical settings, as earlier techniques mainly depended on English datasets and offered vague medication code predictions. MediRec enhances accuracy and transparency by combining clinically relevant reasoning-chain distillation with reinforcement learning. In evaluations against a benchmark for Chinese medication recommendations, it achieved an F1 score of 0.5813. The results were shared in a study on arXiv (arXiv:2510.21084), representing a major advancement in clinical decision-making for Chinese healthcare.
Key facts
- MediRec is an explainable LLM-based framework for Chinese medication recommendation.
- It uses clinically grounded reasoning-chain distillation and reinforcement learning.
- The framework is designed for electronic health records (EHRs).
- Existing approaches are primarily on English datasets and focus on coarse-grained medication code prediction.
- MediRec achieved an F1 score of 0.5813 and a Jaccard score of [incomplete].
- The paper is available on arXiv with ID 2510.21084.
- The announcement type is replace-cross.
- The work aims to improve interpretability in clinical decision-making.
Entities
Institutions
- arXiv