Worst-Case Win Ratios Under Partially Specified Outcome Hierarchies
论文解决了临床试验中部分指定结果层次结构下的最坏情况胜率问题，通过定义估计量为协议允许的比较规则中的最小净收益，并使用U-统计量和大样本理论方法。
0 citationsRead paper
Selected work
Representative Papers
Human-AI Collaboration Increases Efficiency in Regulatory Writing
Drafting nonclinical summaries for Investigational New Drug (IND) applications is time-intensive and heavily reliant on expert knowledge, impeding early-stage drug development efficiency. Method: We propose AutoIND—a human-in-the-loop platform leveraging large language models (LLMs) to automatically generate draft summaries, coupled with a blinded, multidimensional quality assessment framework evaluating seven criteria: correctness, completeness, consistency, clarity, regulatory compliance, traceability, and coherence. Contribution/Results: AutoIND reduces drafting time by 97% (from 100 hours to 2.6–3.7 hours) while processing reports spanning tens of thousands of pages. Generated drafts achieve quality scores of 69.6%–77.9% and contain no critical regulatory deficiencies. This work represents the first systematic application of LLMs to core IND document generation, empirically validating a high-quality, high-efficiency, and regulatory-compliant AI-augmented authoring paradigm. It delivers a reproducible methodology and practical implementation pathway for intelligent drug development.
0 citationsRead paper
Recent publications
Latest Papers
Worst-Case Win Ratios Under Partially Specified Outcome Hierarchies
论文解决了临床试验中部分指定结果层次结构下的最坏情况胜率问题，通过定义估计量为协议允许的比较规则中的最小净收益，并使用U-统计量和大样本理论方法。
0 citationsRead paper
Human-AI Collaboration Increases Efficiency in Regulatory Writing
Drafting nonclinical summaries for Investigational New Drug (IND) applications is time-intensive and heavily reliant on expert knowledge, impeding early-stage drug development efficiency. Method: We propose AutoIND—a human-in-the-loop platform leveraging large language models (LLMs) to automatically generate draft summaries, coupled with a blinded, multidimensional quality assessment framework evaluating seven criteria: correctness, completeness, consistency, clarity, regulatory compliance, traceability, and coherence. Contribution/Results: AutoIND reduces drafting time by 97% (from 100 hours to 2.6–3.7 hours) while processing reports spanning tens of thousands of pages. Generated drafts achieve quality scores of 69.6%–77.9% and contain no critical regulatory deficiencies. This work represents the first systematic application of LLMs to core IND document generation, empirically validating a high-quality, high-efficiency, and regulatory-compliant AI-augmented authoring paradigm. It delivers a reproducible methodology and practical implementation pathway for intelligent drug development.
0 citationsRead paper