Benchmark RadarREMEDI
REMEDI is a benchmark for machine unlearning in multi-label clinical disease inference, built on MIMIC-III, covering diverse forget sets and tasks with utility and unlearning metrics.
- Released
- 2026-06-05
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
It provides a realistic medical-domain evaluation for machine unlearning methods, addressing the lack of benchmarks that reflect real-world patient data and multi-label scenarios, which is crucial for privacy-preserving AI in healthcare.
Motivation
Language models trained for clinical disease inference are trained on patient data, which may include sensitive and private information, and data owners may request the removal of their data from a trained model due to privacy or copyright concerns.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.