English 
ELYZA Inc., which promotes the social implementation of large-scale language models (LLMs), was tasked with developing a domestic specialized LLM platform in the medical field under the research and development task “Development of a Japanese version of a medical LLM and verification of social implementation in clinical settings” in Theme 1 “Research, development and implementation of a medical LLM platform” of the third supplementary budget “Utilization of generative AI in the construction of an integrated healthcare system” of the “Strategic Innovation Promotion Program (SIP)” established by the Cabinet Office. The Japanese version of the medical LLM “ELYZA-LLM-Med” series constructed in this research has achieved the highest performance in Japan in terms of overall score, and has achieved accuracy that exceeds the global top level in on-site medical tasks. In addition, the approach used and verified in the development of this specialized LLM platform is a highly versatile approach for developing specialized models in specific fields, and we believe that it can be horizontally deployed in many fields.
ELYZA will continue to work on cutting-edge research and development, including AI agents, and will support the promotion of social implementation of LLM in Japan and the development of natural language processing technology by making the results of its research publicly available as much as possible.
こちらもお読みください： Mnes’ LOOKREC Integrates with Smart TOHMAS System
Members of this project include Sakura Internet Research Institute, an in-house research institute of Sakura Internet Inc., which oversees the joint research institute as the principal researcher, エリーザ Inc. and the Matsuo-Iwasawa Laboratory at the University of Tokyo, which are responsible for the development of medical-specific LLM, and ABEJA Inc., which is responsible for procuring data, processing it, and developing the LLM system. In addition, when handling personal information, verification was carried out at the RIKEN Institute, which has a secure computing infrastructure environment. For the use cases focused on this time, data was acquired and processed in collaboration with medical institutions such as the International University of Health and Welfare and Fujita Health University, part of the Fujita Gakuen School Corporation.
ソース PRタイムズ
日本語