KuaiRecV2: Benchmarking Large-Scale Continual Learning for Diversified and Multi-task Recommendation | ArxivCSExplorer