a suite of high-quality Chinese datasets, used for pretraining, fine-tuning or preference alignment. And the models trained on these datasets.