Zhihu releases the latest large-scale model application 'Search Aggregation' and starts internal testing today

Zhihu released the latest large-scale model application "Search Aggregation" and started internal testing today. This new feature will provide users with more intelligent and personalized search results, helping users find the information they need faster. PHP editor Strawberry will take you to understand the characteristics and usage of this function, so that you can make better use of this new function and obtain more valuable information.

As previously reported, at the "2023 Zhihu Discovery Conference" in April, Zhihu released the large language model "Zhihaitu AI" and internally tested the "Hot List Summary" of the large model application function on the site. A month later, Zhihu has brought another large model application function "search aggregation" on the site. This product applies large model capabilities to Zhihu search. Whenever a user triggers a search, the system will aggregate opinions from a large number of questions and answers, improving the efficiency of users in obtaining information and forming decisions. Li Dahai said that "search aggregation" will start internal testing today.

At the press conference, Face Wall Intelligence teamed up with the OpenBMB community to open source the self-developed CPM-Bee 10b model. Li Dahai introduced that the model is trained independently from scratch, based on Transformer architecture, excellent bilingual performance in Chinese and English, with tens of billions of parameters and trillions of high-quality corpus. In the ZeroCLUE review, the CPM-Bee 10b scored an overall score of 78.18 On the English common sense knowledge reasoning list, CPM-Bee 10b received an average score of 67 points, which is comparable to the English open source model LLaMA. "CPM-Bee10b It will be fully open source and allowed for commercial use. "Li Dahai said that Wall-Facing Intelligence has always adhered to the open source route and will continue to embrace open source in the future to promote the prosperity of technology and ecology in the field of large models.

The conference also brought the dialogue model product "Luca" developed by Wallface Intelligence. This product has further improved performance on the open source basic model and can perform intelligent interaction and support multiple rounds of dialogue. During the live demonstration at the press conference, "Luca" demonstrated a number of capabilities. It can not only help people understand world knowledge, process mathematical logic, write program codes, and inspire creative inspiration; it can also use massive amounts of knowledge data to help people better obtain Information, planning, problem solving. The press conference also demonstrated the multi-modal understanding ability of "Luca". It can not only analyze picture information such as landscapes and geography, but also understand the emotional meaning conveyed by human pictures. In addition, "Luca" can also find papers and generate abstracts. Li Dahai introduced that "Luka" has now started internal testing.

For more information, please pay attention to this site.

The above is the detailed content of Zhihu releases the latest large-scale model application 'Search Aggregation' and starts internal testing today. For more information, please follow other related articles on the PHP Chinese website!