INDUS, a set of large language models designed for five scientific domains, was developed through a collaboration between NASA and IBM. These models, which include encoders and sentence transformers, were trained using scientific corpora from various data sources. INDUS outperforms non-domain-specific models on various benchmarks and tasks. NASA is incorporating INDUS into its operations to enhance AI capabilities for missions and research. The models have been evaluated and used in different NASA divisions, showing improvements in data categorization, search capabilities, and knowledge retrieval. INDUS is openly available for the scientific community to access and utilize in various applications. An article detailing the capabilities of INDUS is available for further reference.
Article Source
https://www.miragenews.com/nasa-ibm-develop-indus-models-for-advanced-1263004/
