The current AI rails struggling with modern models. It helps to measure the performance of model in certain activities, it can be difficult to know if the internet data models actually solve problems or just remember the answers they have already seen. As models reached 100% in some benches, they also work lowly in revealing meaningful work differences. We continue to invest on new benches and more challenging, but on the road to normal intelligence, we need to look at new ways of testing. The recent replacement of the movement, the test of the judgment person solves these matters memorizing and full memory, but also, it is constantly, it is creating a new difficulty.
While continuing and pursuing existing Ai Benchmarks, we always want to test new ways to test models. That is why today, we introduce KaGle Game Arena: New Platform, Ai Community Public Public Modes where AI models compete in the head of techniques, and give a visible, and powerful.
