نشر الرد: Tencent improves testing originative AI models with unpractical benchmark