xbench

A few days ago, Sequoia China, a well-known investment organization, launched the benchmarking tool "xbench" for AI Agent and published the related technical report. According to the introduction, xbench was created to solve the two core problems of "the relationship between model capability and AI actual utility" and "comparison of capability in different time dimensions". Therefore, Sequoia China put forward the two major directions of "developing new task setting and evaluation methods for real-world utility" and "establishing an evergreen evaluation system". This time, xbench launches a two-track series of assessment sets, which are divided into "xbench-...