Source Bias

Source bias, the undesirable influence of data origin on model outputs, is a significant concern across various machine learning applications, particularly in large language models (LLMs) and image processing. Current research focuses on developing methods to detect and mitigate this bias, employing techniques like parameter-efficient fine-tuning (PEFT) with regularization, counterfactual analysis for bias quantification, and adversarial training to reduce reliance on spurious correlations. Addressing source bias is crucial for ensuring fairness, reliability, and trustworthiness in AI systems, impacting fields ranging from hiring practices to scientific literature analysis and medical image interpretation.