On the Robustness of Transformers against Context Hijacking for Linear Classification

Open in new window