Fetching the paper…

"That Is a Suspicious Reaction!": Interpreting Logits Variation to Detect NLP Adversarial Attacks · Around