SLORR: Simple and Efficient In-Training Low-Rank Regularization
SLORR:简单高效的训练中低秩正则化
David González-Martínez, Shiwei Liu
机构
*
Max Planck Institute for Intelligent Systems(马克斯·普朗克智能系统研究所)
;
University of Tübingen(图宾根大学)
;
ELLIS Institute Tübingen(图宾根ELLIS研究所)
;
Tübingen AI Center(图宾根人工智能中心)
Comments12 pages, 5 figures. has a same-size non-reasoning-teacher control, a three-judge LLM-as-a-judge panel with a negative control, full-source faithfulness grading, and a per-field routing analysis