Learning with noisy labels is a subfield within machine learning and data science focused on developing algorithms and techniques to train robust models effectively using datasets where the assigned labels for training instances contain errors or inaccuracies. This research area investigates methods to mitigate the detrimental effects of label noise on model performance, generalization, and reliability, encompassing approaches such as noise-robust loss functions, noise-aware training strategies, label correction, and noise rate estimation, which is critical for deploying machine learning systems in real-world applications where obtaining perfectly clean labeled data is challenging or cost-prohibitive.