Towards learning to explain with concept bottleneck models: mitigating information leakage

Open in new window