SAES-SVD: Self-Adaptive Suppression of Accumulated and Local Errors for SVD-based LLM Compression

Xing Hu, Dawei Yang, Yuan Cheng, Zhixuan Chen, Zukang Xu · 2026

Read SAES-SVD: Self-Adaptive Suppression of Accumulated and Local Errors for SVD-based LLM Compression by Xing Hu, Dawei Yang, Yuan Cheng, Zhixuan Chen, Zukang Xu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-03051.txt.

Description

Compressing large language models layer by layer ignores how errors pile up, and Xing Hu and 4 co-authors address that blind spot in a 2026 paper. Low-rank shrinking shrinks each layer on its own. Reconstruction error from early layers then propagates and accumulates through the network, amplifying the final deviation from the full-precision model users expected at deployment time.

Their system, SAES-SVD, jointly handles rebuilding within layers and compensation between them. One part uses second-order activation statistics to align each layer's output with the original, explicitly offsetting accumulated error. Another adapts a weighting coefficient to use the rank budget effectively. Tests across architectures and tasks showed consistent gains after shrinking, without fine-tuning or mixed-rank strategies.

Catalog record

CREATORXing Hu, Dawei Yang, Yuan Cheng, Zhixuan Chen, Zukang Xu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE72 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-03051

Browse the full library · About the Tower of Babel