This inventor holds 1 USPTO granted patent. Top assignee: Samsung Electronics Ltd.. Active years: 2026.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile
Company Filing History:
Years Active: 2026
Title: Feixuan Wang: Innovator in Multilingual Language Models
Introduction
Feixuan Wang is a prominent inventor based in Mountain View, CA (US). He has made significant contributions to the field of artificial intelligence, particularly in the development of multilingual large language models. His innovative approach has the potential to enhance communication across different languages.
Latest Patents
Feixuan Wang holds 1 patent for an efficient and effective system and method to build multilingual large language models. This patent describes a method that enables a language model trained in a first language to support a second language. The process includes extending the existing vocabulary of the language model to incorporate additional tokens for text in the second language. Furthermore, it involves initializing these additional tokens based on subtokens of tokens from the first language. The training of the language model utilizes a mixed language dataset that comprises both first and second language corpora. Additionally, the method incorporates instruction tuning using datasets that include instruction and response pairs in both languages.
Career Highlights
Feixuan Wang is currently employed at Samsung Electronics Ltd., where he continues to push the boundaries of language model technology. His work is instrumental in developing systems that can bridge language gaps and improve user experiences in multilingual environments.
Collaborations
Feixuan collaborates with talented individuals such as Hai Wang and Zheng Tang. Their combined expertise fosters an innovative environment that drives advancements in language processing technologies.
Conclusion
Feixuan Wang's contributions to multilingual language models exemplify the intersection of technology and communication. His innovative methods are paving the way for more effective language processing solutions.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile