Provenance-Guided Incremental Learning for Rule-Induced Concept Shift

To handle rule-induced concept shift, this framework uses provenance to trace changes in target-defining rules. It identifies valid historical records through a rule delta and restricts re-labeling to localized regions, ensuring efficient adaptation to evolving definitions.