Post-training alignment of Qwen-based small language models for medication leaflet generation

A study compares post-training methods including supervised fine-tuning (SFT), direct preference optimization (DPO), odds ratio preference optimization (ORPO), and group relative policy optimization (GRPO) on Qwen-based small language models for biomedical data-to-text generation.

• Aligned SLMs outperform proprietary models like GPT-5 in generating medicine package leaflets.
• ORPO outperforms SFT baselines in this specific task.
• GRPO yields the most robust cross-dataset performance among the alignment methods tested as well as GPT-5.

The findings suggest that carefully aligned small language models can achieve superior results compared to larger proprietary alternatives for specialized biomedical translation tasks.