paper-with-me

홈 › Papers

Optimal Brain Decomposition for Accurate LLM Low-Rank Approximation

2026-04-01 · Yuhang Li, Donghyun Lee, Ruokai Yin, Priyadarshini Panda arxiv

Low-rank decomposition has emerged as an important problem in Large Language Model (LLM) fine-tuning and inference. Through Singular Value Decomposition (SVD), the weight matrix can be factorized into low-rank spaces optimally. Previously, a common practice was to decompose the weight in the activation-whitened space, and then achieve satisfying results. In this work, we propose Optimal Brain Decomposition LLM (OBD-LLM), which studies the decomposition problem in the model space by utilizing second-order Hessian information. Through a rigorous Kronecker-factorization of the Hessian, we show that the decomposition needs to consider both input and output information of the layer, and achieves much better decomposition results compared to input only method. Our loss-aware decomposition method involves a bi-directional whitening on the weight matrix. As a result, OBD-LLM is a closed-form solution for the optimal decomposition of weights in the language model. Remarkably, we achieve ~20-40\% better results than previous state-of-the-art decomposition methods, the SVD-LLM.

📄 PDF Abstract BibTeX arXiv:2604.00821

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Scatterbrain: Unifying Sparse and Low-rank Attention Approximation

2021-10-28 · NeurIPS 2021 12 · Beidi Chen, Tri Dao, Eric Winsor, Zhao Song 외

Recent advances in efficient Transformers have exploited either the sparsity or low-rank properties of attention matrices to reduce the computational and memory bottlenecks of modeling long sequences. However, it is stil…

Image GenerationLanguage ModelingLanguage Modelling

Scatterbrain: Unifying Sparse and Low-rank Attention

2021-05-21 · NeurIPS 2021 12 · Beidi Chen, Tri Dao, Eric Winsor, Zhao Song 외

Recent advances in efficient Transformers have exploited either the sparsity or low-rank properties of attention matrices to reduce the computational and memory bottlenecks of modeling long sequences. However, it is stil…

Image GenerationLanguage ModelingLanguage Modelling

A Motion Assessment Method for Reference Stack Selection in Fetal Brain MRI Reconstruction Based on Tensor Rank Approximation

2023-06-30 · Haoan Xu, Wen Shi, Jiwei Sun, Tianshu Zheng 외

Purpose: Slice-to-volume registration and super-resolution reconstruction (SVR-SRR) is commonly used to generate 3D volumes of the fetal brain from 2D stacks of slices acquired in multiple orientations. A critical initia…

MRI ReconstructionSensitivitySuper-Resolution

Low-Rank Dynamic Mode Decomposition: An Exact and Tractable Solution

2016-10-10 · Patrick Héas, Cédric Herzet

This work studies the linear approximation of high-dimensional dynamical systems using low-rank dynamic mode decomposition (DMD). Searching this approximation in a data-driven approach is formalised as attempting to solv…

Robust Randomized Low-Rank Approximation with Row-Wise Outlier Detection

2025-04-03 · Aidan Tiruvan

Robust low-rank approximation under row-wise adversarial corruption can be achieved with a single pass, randomized procedure that detects and removes outlier rows by thresholding their projected norms. We propose a scala…

Outlier Detection