Oppo AI model ranks second to GPT-4 in encyclopedia skills, claims company

Chinese companies are making significant strides in the field of artificial intelligence (AI), with Baidu recently unveiling an industry-grade medical AI model and its ChatGPT competitor, Ernie, capturing substantial public attention. Meanwhile, smartphone manufacturer Oppo has entered the AI chatbot arena and is steadily advancing. Oppo's AI model, known as AndesGPT, has garnered attention for its impressive knowledge and encyclopedia capabilities, ranking second only to OpenAI's GPT-4.

Oppo AI model ranks second to GPT-4 in encyclopedia skills, claims company

Oppo's Xiaobu Assistant proudly announced that AndesGPT secured a place in the September rankings on the SuperCLUE evaluation benchmark. AndesGPT, a generative large language model akin to ChatGPT and Bard, embarked on external testing in early August, alongside Oppo's intelligent assistant, Xiaobu. Their mission was to elevate traditional voice assistants through kernel-level upgrades. Notably, AndesGPT achieved an impressive score of 98.33 in the "knowledge and encyclopedia" capability test, claiming the second spot globally, with the top position in China, trailing only behind GPT-4.

Oppo disclosed that it has significantly enhanced AndesGPT's knowledge and encyclopedic abilities. This was achieved by integrating high-quality knowledge graphs, a substantial repository of question-and-answer data, and optimizing pre-training and fine-tuning techniques.

Last month, Oppo unveiled plans to launch a large-scale experiential campaign for its Xiaobu Assistant, which relies on the foundation of the AndesGPT large language model. This advanced AI assistant is equipped to generate tailored content based on user inputs and boasts remarkable AI capabilities in summarization and synthesis, promising substantial improvements in these areas.