Huawei Xiaoyi lab details its path toward proactive multimodal assistants
In an interview with LeiPhone AI, Liu Rui, founding director of Huawei’s Xiaoyi Large Model Application Lab, described the lab’s work on moving AI assistants from passive, question-and-answer interactions toward proactive services. He presented Aura, a persistent assistant designed to process continuous visual and audio context, along with the open-source image-generation model Boogu, the Math Canvas visual-reasoning project, the Omni-Interact evaluation benchmark, and agent-training work including GHPO and COVE. Claims about Boogu’s generation quality and training cost, as well as the lab’s other results, came from Liu’s account of the research.
The interview provides specific details about Huawei Xiaoyi’s research into persistent multimodal interaction, image generation, visual reasoning and agent evaluation, while making clear that the reported results came from the lab director.
导语：在商汤没做成的技术，登上了华为发布会。
在商汤没做成的技术，登上了华为发布会。 作者：李秋悦 编辑：吴彤
2022年6月，刘睿港中文博士毕业，赶上的却是一个“坏时候”。
深度学习的高光已持续近十年，新突破越来越难；疫情之下，经济低迷，同届不少同学选择延期，再守一年实验室。
刘睿没有等，他直接去了商汤。
#Huawei#Xiaoyi Large Model Application Lab#SenseTime#Aura#Boogu#Math Canvas#Omni-Interact#GHPO#COVE#Pangu