1 paper
Tuan Le, Wei Qian, Mengdi Huai
Large Reasoning Models (LRMs) generate structured chains of thought (CoTs) before producing final answers, making them especially vulnerable to knowledge leakage through intermedia…