Large Reasoning Models (LRMs) have achieved impressive performance via explicit Chain-of-Thought (CoT) reasoning, yet this process introduces critical safety…
机构:浙大
来源:arXiv 2611.07482 | AI4Papers 论文推荐平台