CORTEXA
← Browse
arxivcs.LGcs.CV2026-07-05

SAD-LoRA: Spectral Alignment for Low-Rank Knowledge Distillation

Omer Tariq, Syed Muhammad Raza, Jeongbae Son

Distilling a fine-tuned teacher into a LoRA-adapted student is a standard recipe for parameter-efficient compression, but output-level KD does not explicitly control which rank-$r$ weight subspace the adapter occupies. We propose \textbf{SAD-LoRA} (\textbf{S}pectral \textbf{A}lignment \textbf{D}istillation), which selects this subspace from the data-weighted student-space reference update $\DWT\Sigx^{1/2}$ and maintains it during training via a differentiable principal-angle loss on $\colspan(B)$. We show that the data-weighted distillation error decomposes exactly into subspace misalignment, within-subspace coefficient mismatch, and irreducible rank residual; standard KD can affect the first term only indirectly through output gradients. On controlled synthetic problems with a flat teacher spectrum, SAD-LoRA reduces the subspace-misalignment term from $51\%$ to nearly zero and lifts final subspace alignment from $0.49$ to $1.00$. On RoBERTa-large to RoBERTa-base distillation across six GLUE tasks, SAD-LoRA improves rank efficiency: at $r{=}4$, it matches or beats the strongest included spectral baseline on five of six tasks, and at $r{=}8$ it gives the best result on SST-2 and CoLA. Ablations identify subspace alignment as the load-bearing component, while coefficient matching is auxiliary.

View free PDFSource page

Related papers

arxivcs.CVcs.AIcs.LG2026-07-08

LoCA: Spatially-Aware Low-Rank Convolutional Adaptation of Vision Foundation Models

Sojung An, Junha Lee, Sujeong You, Nam Ik Cho, Donghyun Kim

Pre-trained Vision Foundation Models (VFMs) provide strong visual representations for diverse downstream tasks. The key challenge of VFM adaptation stems from the prohibitive costs of full fine-tuning and catastrophic forgetting. To address this, Low-Rank Adaptation (LoRA) has em…

View free PDFSource page
arxivcs.CVcs.LG2026-07-18

Transferable Low-Rank Convolutional Bases for Onboarding Unseen Medical Imaging Modalities

Ranat Das Prangon, Istiaque Ahmed, Shajid Hasan Naim, Waseem Mustak Zisan, Hossain Md Shakhawat

Deploying a medical imaging model that must later accommodate a modality it has never seen is a recurring practical problem: retraining the shared representation is expensive and destroys performance on the modalities already in service. We study this \emph{onboarding} problem un…

View free PDFSource page
arxivcs.CVcs.AIcs.LG2026-07-05

Hierarchical Multi-to-Single-Modal Knowledge Distillation for Disruption Prediction in EAST

Qiang Chen, Xiao Wang, Hao Si, Qingquan Yang, Meiwen Chen, Jianhua Yang, et al.

Plasma disruption is a critical threat to tokamak safety. Existing data-driven predictors mainly rely on time-series diagnostic signals, while visible images provide complementary spatial cues including plasma deformation, local brightening, and radiation-structure evolution. Alt…

View free PDFSource page
arxivcs.CVcs.AIcs.LG2026-07-07

Few-Medoids: An Embarrassingly Simple Coreset Selection Method for Few-Shot Knowledge Distillation

Cemil-Andrei Dilmac, Florinel-Alin Croitoru, Radu Tudor Ionescu

Coreset selection aims to identify a small and highly representative subset of a massive dataset for efficient model training. The problem remains challenging even in the few-shot knowledge distillation (KD) setup, where a full-scale pre-trained teacher informs the student networ…

View free PDFSource page
arxivcs.GRcs.AIcs.CVcs.DCcs.LG2026-06-30

Benchmarking Federated Learning and Knowledge Distillation for Point Cloud Classification

Aizierjiang Aiersilan

Deploying 3D point cloud analysis in privacy-sensitive, resource-constrained settings faces two barriers: data cannot be centralized, and models must run on limited edge hardware. We present a multi-seed benchmark jointly evaluating federated learning (FL) and knowledge distillat…

View free PDFSource page