A comparative evaluation of chain-of-thought-based prompt engineering techniques for medical question answering.
基於 chain-of-thought 的提示工程技術於醫學問答之比較性評估
Comput Biol Med 2025-07-02
OpenMedLM: prompt engineering can out-perform fine-tuning in medical question-answering with open-source large language models.
OpenMedLM: 在醫學問答中,prompt engineering 可以優於對開源大型語言模型進行微調。
Sci Rep 2024-06-19
Evaluation of LLMs accuracy and consistency in the registered dietitian exam through prompt engineering and knowledge retrieval.
透過提示工程和知識檢索評估大型語言模型在註冊營養師考試中的準確性和一致性。
Sci Rep 2025-01-09
Empowering large language models for automated clinical assessment with generation-augmented retrieval and hierarchical chain-of-thought.
利用生成增強檢索和分層思維鏈來提升大型語言模型的自動臨床評估能力。
Artif Intell Med 2025-02-20
Role of Model Size and Prompting Strategies in Extracting Labels from Free-Text Radiology Reports with Open-Source Large Language Models.
開源大型語言模型在從自由文本放射學報告中擷取標籤時,模型規模與提示策略的角色
J Imaging Inform Med 2025-05-05
開源大型語言模型(LLMs)在從放射科報告擷取標籤上,比傳統規則式工具(如 CheXpert)更準確。規模較大的 LLMs,特別在判讀困難異常(如肋骨骨折)時,敏感度更高。不同提示策略(如 chain-of-thought)效果不一。即使標籤有雜訊,用 LLM 擷取的標籤訓練影像分類器,表現仍不錯,但最終評估結果會受標註方法影響很大。因此,選對 LLM、提示方式和評估方法對醫療 AI 發展很重要。
PubMedDOI
Cross-Institutional Evaluation of Large Language Models for Radiology Diagnosis Extraction: A Prompt-Engineering Perspective.
跨機構評估大型語言模型於放射診斷萃取之表現:以提示工程觀點分析
J Imaging Inform Med 2025-05-09
Chain of Thought Strategy for Smaller LLMs for Medical Reasoning.
用於醫學推理之較小型LLM的Chain of Thought策略
Stud Health Technol Inform 2025-05-17
這篇論文發現,用 Chain of Thought(CoT)提示法能讓小型語言模型在醫學問答上表現更好、更透明,特別是在 PubMedQA 資料集上效果明顯。CoT 幫助模型逐步推理,提升準確度和可解釋性。不過,遇到很專業的題目還是有困難。若結合檢索增強生成等技術,小型模型未來有機會追上大型模型。
PubMedDOI
Evaluating performance of large language models for atrial fibrillation management using different prompting strategies and languages.
使用不同提示策略與語言評估大型語言模型於心房顫動管理的表現
Sci Rep 2025-05-30
Evaluating large language models for information extraction from gastroscopy and colonoscopy reports through multi-strategy prompting.
透過多策略提示評估大型語言模型於胃鏡與大腸鏡報告資訊擷取的表現
J Biomed Inform 2025-06-12