Comparing the performance of four mainstream large language models on medical literature review generation: a human expert evaluation in SMILE surgery
Rather than excelling universally, LLMs exhibit distinct and task-specific strengths that mandate a task-driven, hybrid strategy in tool selection. Reference fabrication was found to be a pervasive issue across all models, regardless of the task topic, elevating human verification from a best practice to an essential safeguard for academic integrity.
Authors
- Chuyang Lin
- Fu Gui (ORCID: https://orcid.org/0000-0001-8329-7118)
- Wenlei Liu
- Bin Zhao (ORCID: https://orcid.org/0000-0002-6974-2323)
- Jiayu Li (ORCID: https://orcid.org/0000-0002-2755-6549)
- Mengyun Zhou
- Zeyu Zhu
- Chong Ai
- Jian Xiong
- Jie Liu
- Fei Huang
- Xian Zhang
- Fei Zhong
- Dongmei Han
- Ling Ling
- Yalin Lu
Institutions
- Nanchang University (CN)
- Sun Yat-sen University (CN)
- Shanghai Eye Disease Prevention & Treatment Center (CN)
- ShenZhen People’s Hospital (CN)
- The Central Hospital of Xiao gan (CN)
- Xiaogan First People's Hospital (CN)
- Second Affiliated Hospital of Nanchang University (CN)
- Eye & ENT Hospital of Fudan University (CN)
- Third Hospital of Nanchang (CN)
- First Affiliated Hospital of Xi'an Jiaotong University (CN)
- First Affiliated Hospital of Gannan Medical University (CN)
- He Eye Hospital (CN)
- First Hospital of Xi'an (CN)
Publication Details
- Journal
- Graefe s Archive for Clinical and Experimental Ophthalmology
- Published
- 2026-01-03
- DOI
- https://doi.org/10.1007/s00417-025-07092-1
- Primary Topic
- Artificial Intelligence in Healthcare and Education
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China
- Health Commission of Jiangxi Province