Weak supervision has emerged as a powerful technique for training machine learning models, leveraging noisy or readily available data sources. However, ensuring the reliability of weakly supervised labels remains a significant challenge. RWIn presents a novel framework designed