Adaptive Distraction: Probing LLM Contextual Robustness with Automated Tree Search | Read Paper on Bytez