arXiv Computation and Language

Garbage Attention in Large Language Models: BOS Sink Heads and Sink-aware Pruning

arXiv Machine Learning
Aug 27

When Pruning Meets Interpretability: Preserving Sparse Autoencoder Robustness in LLMs

Sparse autoencoders (SAEs) are commonly used to interpret large language models, but their reliability after pruning is unclear. This study shows that pruning’s effect on an SAE is governed by perturbation energy, a covariance-weighted norm, and that magnitude pruning distorts the representation space by ignoring activation geometry. Activation-aware pruning methods such as Wanda and SparseGPT better preserve SAE behavior, and the authors find that middle layers are especially vulnerable, leading them to propose a layer‑wise sparsity allocation that reduces perplexity for a given sparsity level.

By Suchit Gupte, Xueru Zhang, Mohammad Mahdi Khalili
arXiv AI
Nov 6

When Bigger Becomes Brittle: Why Compression Hurts More as Language Models Scale

The paper investigates how quantization and pruning—common compression techniques—affect language models of different sizes. By isolating a small set of attention heads responsible for in-context retrieval, the authors decompose the loss from ablating these heads into components that are compensated by the rest of the network and components that directly impact the output. Their findings show that while smaller models largely recover from such damage, larger models retain most of the loss because their compensatory mechanisms are less effective, indicating that compression can be far more detrimental as models scale.

By Gautam Gupta, Dikshant Kukreja, Unique Kumar, Avni Mittal, Deepanshu Dabas, Kshitij Sah, Erik Cambria, Rajiv Ratn Shah, Zhengkui Wang, Avinash Anand
Hugging Face Trending Papers
Nov 5

When Bigger Becomes Brittle: Why Compression Hurts More as Language Models Scale

The paper investigates how quantization and pruning affect language models of different sizes. It shows that compression damages specific components—particularly attention heads used for in‑context retrieval—more severely in larger models, and that the network’s ability to compensate for this damage diminishes as model size grows. Consequently, compression techniques that appear benign on small models can become costly on larger ones.

arXiv AI
Oct 2

Sequential Functional Structured Tucker Compression for Large Language Model Attentions

The paper introduces FTC, a sequential functional structured Tucker compression method for large language model attentions. FTC adapts its approximation to the current compressed model, jointly exploits the Q/K/V head structure, and treats the output projection separately to handle representation shifts. It achieves the lowest WikiText‑2 perplexity across seven decoder‑only LLMs (6B–32B) at all tested keep ratios, with significant gains under aggressive compression and benefits that transfer to downstream tasks.

By Jiangfeng Chen, Xinyu Wang, Tianshuo Yan, Hanwei Wu, Xiao-Wen Chang, Yang Zhang, Lei Ding