研究人员开发了新的方法，可以在推理时引导大型语言模型（LLM）的行为，而不会牺牲生成质量。一种方法是仅提示词转向向量（Prompt-only SV, PrOSV），它仅干预提示词标记，在AxBench等基准测试中表现优于传统的全序列转向向量。另一种方法是FLAS（基于流的激活转向），它学习一个条件概念的速度场来传输激活，在Gemma模型上始终优于提示词方法。第三种技术SKOP（通过关键正交投影进行转向）将注意力重新路由限制在保留推理和检索性能上，在效用和转向效果之间取得了更好的权衡。 AI
影响 推理时LLM控制的新技术可以通过提高转向准确性和减少性能下降，从而实现更细致、更可靠的AI应用。
排序理由 三篇新的arXiv论文介绍了在推理时控制LLM行为的新颖方法。
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
arXiv:2605.05983v1 Announce Type: new Abstract: Recently, steering vectors (SVs) have emerged as an effective and lightweight approach to steer behaviors of large language models (LLMs), among which fine-tuned SVs are more effective than optimization-free ones. However, current a…
arXiv:2605.05892v1 Announce Type: cross Abstract: Activation steering has emerged as a promising alternative for controlling language-model behavior at inference time by modifying intermediate representations while keeping model parameters frozen. However, large-scale evaluations…
arXiv:2605.06342v1 Announce Type: new Abstract: Activation steering controls LLM behaviour towards target behaviour by intervening in internal representations, yet it often degrades reasoning and retrieval performance. We argue that a primary cause of this trade-off is attention …
Activation steering controls LLM behaviour towards target behaviour by intervening in internal representations, yet it often degrades reasoning and retrieval performance. We argue that a primary cause of this trade-off is attention rerouting: steering vectors alter query-key matc…
Activation steering has emerged as a promising alternative for controlling language-model behavior at inference time by modifying intermediate representations while keeping model parameters frozen. However, large-scale evaluations such as AxBench show that existing steering metho…