Train Once, Reuse Everywhere: Generalizable Implicit In-Context Learning by Routing Attention 事件

Train Once, Reuse Everywhere: Generalizable Implicit In-Context Learning by Routing Attention arXiv:2509.22854v2 Announce Type: replace Abstract: Implicit in-context learning (ICL) has newly emerged as a promising paradigm that simulates ICL behaviors in the representation space of large language models (LLMs), aiming to attain few-shot performance at zero-shot cost. However, existing approaches largely rely on injecting shift vectors into residual flows, which are typically constructed from la