Risk Under Pressure: Compute-Aware Evaluation of Adversarial Robustness in Language Models | Malikeh Ehghaghi et al. | ResearchPod