Label Bias

Label bias, the systematic skew in training data labels due to various factors like human biases or sampling limitations, significantly impacts the fairness and accuracy of machine learning models across diverse applications. Current research focuses on quantifying and mitigating this bias in various model architectures, including large language models (LLMs), graph neural networks (GNNs), and image classifiers, employing techniques such as expectation-maximization, confident learning, and causal inference methods to improve model robustness and fairness. Addressing label bias is crucial for ensuring the reliability and ethical deployment of machine learning systems, particularly in sensitive domains like healthcare and criminal justice, where biased predictions can have severe consequences.