Fetching the paper…

TASA: Deceiving Question Answering Models by Twin Answer Sentences Attack · Around