WenXinGPT: A Multimodal Conversational Model for Enhancing Orthopedic Expert Consultations
Keywords: Multimodal conversational model, orthopedic expert consultations, medical visual language model, zero-shot scenarios, large language models
Abstract: Inspired by the hospital expert consultation model, this paper proposes a conversational medical visual language model for orthopedics, named WenXinGPT (Multi-disciplinary Collaboration). The core concept of this work focuses on aligning medical visual and textual representations to leverage high-quality data for generating expert consultation dialogues across hospital departments. The primary objective is to uncover orthopedic knowledge within medical intelligence models and enhance their reasoning abilities in an interpretable manner without requiring additional training. Our research particularly emphasizes zero-shot scenarios, and the results from experiments on 16 datasets provided by Peking Union Medical College Hospital demonstrate that the proposed WenXinGPT framework excels at mining and utilizing medical expertise within large language models, while also expanding their reasoning capabilities. Based on these findings, we conducted manual evaluations to identify and categorize common errors in our methods, along with ablation studies aimed at understanding the impact of various factors on overall performance.
Primary Area: foundation or frontier models, including LLMs
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5190
Loading