不只是拒绝请求，更是防住有害动作：这项研究揭示安全训练在智能体环境中能否真正“扛住”优化冲刷。
arXiv:2603.02229v2 Announce Type: replace Abstract: Safety post-training has been studied extensively in single-step "chat" settings where safety typi…
揭示大模型安全训练中“过度拒绝”的成因，用竞争奖励机制平衡安全与可用性。
arXiv:2606.31748v1 Announce Type: new Abstract: Safety training on language models often induces over-refusal: improved safety on harmful prompts at t…
研究发现对齐训练并非消除文化知识，而是通过抑制机制在拒绝时刻压制模型内部logit分布。
arXiv:2510.13154v2 Announce Type: replace Abstract: What happens inside a language model when alignment training conflicts with a cultural value it en…