Cascading Adversarial Bias from Injection to Distillation in Language Models Paper • 2505.24842 • Published May 30, 2025 • 6