Exploring Post-Training Alignment of Small Language Models for Biomedical Data-to-Text Generation: A Case Study of Medication Leaflet
For biomedical informatics practitioners, this work provides a comparative evaluation of alignment methods for small models in a specialized generation task, but the findings are incremental and dataset-specific.
This study compares post-training alignment methods (SFT, DPO, ORPO, GRPO) for small language models on biomedical data-to-text generation using medication leaflet and drug label datasets. Results show that aligned SLMs outperform GPT-5, with ORPO beating SFT baselines and GRPO achieving the most robust cross-dataset performance.
Translating complex biomedical data into patient-friendly narratives is central to modern biomedical informatics. This study presents a comparative analysis of training small language models (SLMs) in specialized biomedical datato-text generation tasks. We explore widely adopted post-training methods including supervised fine-tuning (SFT), direct preference optimization (DPO), odds ratio preference optimization (ORPO), and group relative policy optimization (GRPO) with Qwen-based SLMs on a medicine package leaflets dataset. To assess cross-dataset generalizability, we also curated drug label data from openFDA. We evaluate models using both standard lexical overlap metrics like ROUGE as well as semantic similarity measures. Across our experiments, the results show that (1) the aligned SLMs outperform proprietary models like GPT-5; (2) ORPO outperforms the SFTbaselines; (3) GRPO yields the most robust cross-dataset performance among the alignment methods tested as well as GPT-5.