Clinical named entity recognition from dental progress notes is challenging because documentation is highly unstructured, domain-specific, and often privacy-sensitive. We developed a locally deployable framework that enables small language models to self-generate, verify, refine, and evaluate entity-specific prompts for extracting multiple clinical entities from dental notes. Using 1,200 annotated notes, we evaluated candidate open-weight models with multi-prompt ensemble inference and further adapted selected models using QLoRA-based supervised fine-tuning and direct preference optimization. Model performance varied substantially, highlighting the need for task-specific evaluation rather than reliance on generic benchmarks. Qwen2.5-14B-Instruct achieved the strongest baseline performance. After DPO, Qwen2.5-14B-Instruct and Llama-3.1-8B-Instruct achieved micro/macro F1 scores of 0.864/0.837 and 0.806/0.797, respectively. These findings suggest that automated prompt optimization combined with lightweight preference-based post-training can support scalable clinical information extraction using locally deployed small language models.

翻译：牙科病程记录的临床命名实体识别具有高度非结构化、领域特异性强及隐私敏感性高的特点，这使得该任务面临诸多挑战。我们开发了一种可本地部署的框架，使小语言模型能够自主生成、验证、优化及评估面向实体的提示，从而从牙科记录中抽取多种临床实体。基于1200条已标注记录，我们采用多提示集成推理对候选开源模型进行评估，并进一步通过基于QLoRA的监督微调与直接偏好优化对所选模型进行适配。模型性能差异显著，这表明需开展任务特定评估而非依赖通用基准。Qwen2.5-14B-Instruct取得了最优基线性能。经DPO优化后，Qwen2.5-14B-Instruct与Llama-3.1-8B-Instruct的宏/微平均F1值分别达到0.864/0.837及0.806/0.797。研究结果表明，自动提示优化结合轻量级偏好后训练能够支撑基于本地部署小语言模型的可扩展临床信息抽取。