Text ranking models based on BERT are now well established for a wide range of pas- sage and document ranking tasks. However, the robustness of BERT-based ranking models under adversarial attack is under-explored. In ...