Oliver Daniels comments on Latent Adversarial Training

[Oliver Daniels]

Is anyone aware of preliminary empirical work here? (Not including standard adversarial training)