让模型学习到与任务相关的独特特征方向，提升表征可解释性。
Objective-Specific Privileged Bases via Full-Prefix Matryoshka Learning
- 通过全前缀马特罗什卡学习，构建任务导向的特征基
- 各维度大小反映信息量，与任务信号高度对齐
- 适用于需要可解释特征表示的下游任务
学习到的表征通常对旋转变换不变，导致各维度不可区分且可互换。我们研究马特罗什卡表征学习（MRL）如何生成不同于方差或正则化引导的、与任务对齐的特权基。在线性设置下，证明全前缀MRL能恢复有序主方向，并可通过共享统计量高效计算。实证表明，MRL产生的每维结构与任务信号保持一致，坐标幅值反映其信息量。
原文摘要 · Abstract (English)
Learned representations are often invariant to rotational transformations, leaving individual dimensions non-identifiable and interchangeable. We study how Matryoshka Representation Learning (MRL) induces a task-aligned privileged basis distinct from variance-based or regularizer-induced orderings. In the linear setting, we prove that full-prefix MRL recovers the ordered principal directions, and can be computed efficiently using shared statistics. Empirically, we demonstrate that MRL yields consistent per-dimension structure aligned with task signal, where coordinate magnitude reflects informativeness.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。