一项新研究调查了加性激活引导从单轮聊天到ReAct代理的可迁移性，发现虽然引导方向一致地到达了后期层，但其行为影响是不可预测且依赖于模型的。研究表明，在某些模型上，代理部署可以将拒绝绕过向量放大高达2.00倍，而其他模型则显示出衰减，这表明安全不能被假定。这种分离表明重缩放发生在ReAct格式脚手架上，而不是工具观察上。 AI
影响 加性激活引导在代理AI部署中不可预测的安全结果，需要仔细的模型选择和安全评估。
排序理由 学术论文，详细介绍了一项关于AI模型行为和安全影响的新研究。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2607.09156v1 Announce Type: new Abstract: Additive activation steering (injecting a scaled residual-stream direction during generation) is calibrated almost entirely in single-turn chat, yet the models it targets are increasingly deployed as tool-using ReAct agents. We pres…
Additive activation steering (injecting a scaled residual-stream direction during generation) is calibrated almost entirely in single-turn chat, yet the models it targets are increasingly deployed as tool-using ReAct agents. We present the first systematic chat-to-agent transfer …