Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yu, Hongzhou, Cheng, Tianhao, Wang, Yingwen, He, Wen, Wang, Qing, Cheng, Ying, Zhang, Yuejie, Feng, Rui, Zhang, Xiaobo
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2501.09213
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912509948067840
author Yu, Hongzhou
Cheng, Tianhao
Wang, Yingwen
He, Wen
Wang, Qing
Cheng, Ying
Zhang, Yuejie
Feng, Rui
Zhang, Xiaobo
author_facet Yu, Hongzhou
Cheng, Tianhao
Wang, Yingwen
He, Wen
Wang, Qing
Cheng, Ying
Zhang, Yuejie
Feng, Rui
Zhang, Xiaobo
contents Recent advancements in large language models (LLMs) have shown promise in medical applications such as disease diagnosis and treatment planning. However, most existing medical LLMs struggle with the deep reasoning required for complex medical problems, such as differential diagnosis and medication recommendations. We propose FineMedLM-o1, which leverages high-quality medical synthetic data and long-form reasoning data for Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO), enabling advanced dialogue and deep reasoning capabilities. Additionally, we introduce Test-Time Training (TTT) in the medical domain for the first time, facilitating domain adaptation and ensuring reliable, accurate reasoning. Experimental results demonstrate that FineMedLM-o1 achieves a 23% average performance improvement over prior models on key medical benchmarks. Furthermore, the introduction of TTT provides an additional 14% performance boost, highlighting its effectiveness in enhancing medical reasoning capabilities. To support this process, we also propose a novel method for synthesizing medical dialogue. Compared to other open-source datasets, our dataset stands out as superior in both quality and complexity. The project and data will be released on GitHub.
format Preprint
id arxiv_https___arxiv_org_abs_2501_09213
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FineMedLM-o1: Enhancing Medical Knowledge Reasoning Ability of LLM from Supervised Fine-Tuning to Test-Time Training
Yu, Hongzhou
Cheng, Tianhao
Wang, Yingwen
He, Wen
Wang, Qing
Cheng, Ying
Zhang, Yuejie
Feng, Rui
Zhang, Xiaobo
Computation and Language
Recent advancements in large language models (LLMs) have shown promise in medical applications such as disease diagnosis and treatment planning. However, most existing medical LLMs struggle with the deep reasoning required for complex medical problems, such as differential diagnosis and medication recommendations. We propose FineMedLM-o1, which leverages high-quality medical synthetic data and long-form reasoning data for Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO), enabling advanced dialogue and deep reasoning capabilities. Additionally, we introduce Test-Time Training (TTT) in the medical domain for the first time, facilitating domain adaptation and ensuring reliable, accurate reasoning. Experimental results demonstrate that FineMedLM-o1 achieves a 23% average performance improvement over prior models on key medical benchmarks. Furthermore, the introduction of TTT provides an additional 14% performance boost, highlighting its effectiveness in enhancing medical reasoning capabilities. To support this process, we also propose a novel method for synthesizing medical dialogue. Compared to other open-source datasets, our dataset stands out as superior in both quality and complexity. The project and data will be released on GitHub.
title FineMedLM-o1: Enhancing Medical Knowledge Reasoning Ability of LLM from Supervised Fine-Tuning to Test-Time Training
topic Computation and Language
url https://arxiv.org/abs/2501.09213