On September 7th, news emerged that the 'Fangsheng' agent benchmark, a pioneering initiative by the China Academy of Information and Communications Technology (CAICT), has developed a multi-tiered testing framework. This framework systematically evaluates agents across three dimensions: fundamental capabilities, practical application scenarios, and intricate integrated tasks. Its core objective is to guide agent testing towards a holistic assessment of capabilities, in-depth process analysis, and precise problem identification.

In a bid to elevate the benchmark's professionalism, enhance its representativeness, and ensure its alignment with industry needs, CAICT has officially opened a public call for test task submissions for the 'Fangsheng' agent benchmark. This initiative aims to expand the repository of high-quality test tasks, while also refining the testing tasks themselves, the testing environments, and the evaluation methodologies to better serve the industry's evolving requirements.