Exploiting contrastive learning to learn a good representation becomes one of the promising fundamental tasks in unsupervised learning. The notorious problem in contrastive learning, where large batch sizes and long training epochs are required to train the model, is intractable. Nevertheless, with the increment of the batch sizes, the proportion of false negatives in each batch also increases, leading to a diminishment of model performance. We rigorously demonstrate the impact of false negatives on the model’s representational capacity and propose a credit-based negative sample denoising technique. We conduct experiments on two datasets, CIFAR-10 and Fashion-MNIST, with image classification as the downstream task. Compared to current traditional methods, our approach achieved better performance on the original baseline.