Return to Article Details Embedding Instability Score for Detecting Backdoor Attacks in NLP Models Download Download PDF