Cascading Adversarial Bias from Injection to Distillation in Language Models | Read Paper on Bytez