ConceptConcept Erasure
Concept Erasure
A technique for removing specific sensitive information (such as gender or race) from the representations of a machine learning model while preserving task-relevant information, with the aim of achieving fairness and privacy protection.
Timeline (last 90 days)
Nov 10
For practitioners working on interpretability and fairness of multilingual models, one should not assume that a concept erasure method effective in one language can transfer directly to another language.
Unverified50%