New Framework Automates Data Labeling with 98.9% Coverage, Reducing Manual Annotation Costs

Source: ArXiv CS.LG. Not independently corroborated. Researchers introduced EXPONA, an automated framework that generates reliable label functions for machine learning training data without manual annotation. The system explores multiple labeling approaches while filtering out noisy rules, achieving near-complete coverage and 46% performance gains across diverse datasets—addressing a critical bottleneck in AI model development.