Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing 文章

ArXiv CS.CL2026-05-28NEWSen作者: Li Lei, Madalina Ciobanu, Qingqing Mao, Ritankar Das
查看原文 →
关系图谱

Interpretability-Guided Layer Selection over Subspace Projection: SAEs as Stethoscopes, Not Scalpels, for Raw Task Vector Model Editing · 相关技术

相关技术

sparse autoencodertask vectors