A lightweight large language model for regulatory affairs translation in pharmaceutical industry
8.4
来源：
Nature
关键字：
AI drug discovery
发布时间：
2025-10-31 03:53
摘要：
PhT-LM is a novel lightweight large language model developed to enhance the translation of regulatory affairs documents in the pharmaceutical industry. By utilizing advanced techniques such as retrieval-augmented generation, it significantly improves translation quality and reduces costs associated with traditional translation methods. The model has shown superior performance in both English-Chinese and Chinese-English translation tasks, making it a valuable tool for pharmaceutical companies navigating complex regulatory environments.
原文：
查看原文
价值分投票
评分标准
新闻价值分采用0-10分制，综合考虑新闻的真实性、重要性、时效性、影响力等多个维度。
评分越高，表示该新闻的价值越大，越值得关注。
价值维度分析
domain_focus
1.0分+1.0分+重点关注领域符合度
business_impact
1.0分+商业影响力
scientific_rigor
1.5分+具体实验数据支持
timeliness_innovation
1.5分+技术创新
investment_perspective
2.5分+早期研发阶段
market_value_relevance
1.0分+市场价值相关性
team_institution_background
0.5分+团队背景
technical_barrier_competition
1.0分+技术壁垒
关键证据
PhT-LM achieved a BLEU-4 mean score of 36.018 and a CHRF mean score of 58.047.
The model significantly reduces translation costs and improves efficiency.
It is designed specifically for regulatory affairs, addressing industry-specific challenges.
真实性检查
否
AI评分总结
PhT-LM is a novel lightweight large language model developed to enhance the translation of regulatory affairs documents in the pharmaceutical industry. By utilizing advanced techniques such as retrieval-augmented generation, it significantly improves translation quality and reduces costs associated with traditional translation methods. The model has shown superior performance in both English-Chinese and Chinese-English translation tasks, making it a valuable tool for pharmaceutical companies navigating complex regulatory environments.