Saved in:
Bibliographic Details
Main Authors: Chen, Lei, Meng, Yuan, Zhan, Xiaoyu, Wang, Zhi, Zhu, Wenwu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2602.14452
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910023458750464
author Chen, Lei
Meng, Yuan
Zhan, Xiaoyu
Wang, Zhi
Zhu, Wenwu
author_facet Chen, Lei
Meng, Yuan
Zhan, Xiaoyu
Wang, Zhi
Zhu, Wenwu
contents Large Language Models (LLMs) offer strong capabilities but incur high inference costs due to dense computation and memory access. Training-free activation sparsity is a promising approach for efficient LLM inference, yet existing methods often rely solely on activation information and uniform sparsity ratios. This overlooks the critical interplay with weights and inter-block sensitivity variation, leading to suboptimal performance. We identify two key phenomena in modern LLMs: 1) less significant activations may align with highly important weights, and 2) sparsity sensitivity varies non-monotonically across model blocks. We propose Weight-aware Mixed-Granularity Training-free Activation Sparsity (WiSparse), which leverages both activation and weight information for adaptive sparsity allocation. Specifically, we introduce a weight-aware mechanism integrating activation magnitudes with precomputed weight norms to accurately identify salient channels. This is combined with a mixed-granularity allocation scheme: a global budget is distributed across blocks via evolutionary search to protect sensitive regions, then refined within blocks to minimize reconstruction error. We improve sparse kernels and demonstrate effectiveness on three representative models. Notably, at 50% sparsity, WiSparse preserves 97% of Llama3.1's dense performance, surpassing the strongest baseline by 2.23 percentage points while achieving a 21.4% acceleration in end-to-end inference speed. Our research advances the limits of training-free approaches for efficient LLM inference, pushing the boundaries of achievable speedup without training.
format Preprint
id arxiv_https___arxiv_org_abs_2602_14452
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle WiSparse: Boosting LLM Inference Efficiency with Weight-Aware Mixed Activation Sparsity
Chen, Lei
Meng, Yuan
Zhan, Xiaoyu
Wang, Zhi
Zhu, Wenwu
Machine Learning
Artificial Intelligence
Large Language Models (LLMs) offer strong capabilities but incur high inference costs due to dense computation and memory access. Training-free activation sparsity is a promising approach for efficient LLM inference, yet existing methods often rely solely on activation information and uniform sparsity ratios. This overlooks the critical interplay with weights and inter-block sensitivity variation, leading to suboptimal performance. We identify two key phenomena in modern LLMs: 1) less significant activations may align with highly important weights, and 2) sparsity sensitivity varies non-monotonically across model blocks. We propose Weight-aware Mixed-Granularity Training-free Activation Sparsity (WiSparse), which leverages both activation and weight information for adaptive sparsity allocation. Specifically, we introduce a weight-aware mechanism integrating activation magnitudes with precomputed weight norms to accurately identify salient channels. This is combined with a mixed-granularity allocation scheme: a global budget is distributed across blocks via evolutionary search to protect sensitive regions, then refined within blocks to minimize reconstruction error. We improve sparse kernels and demonstrate effectiveness on three representative models. Notably, at 50% sparsity, WiSparse preserves 97% of Llama3.1's dense performance, surpassing the strongest baseline by 2.23 percentage points while achieving a 21.4% acceleration in end-to-end inference speed. Our research advances the limits of training-free approaches for efficient LLM inference, pushing the boundaries of achievable speedup without training.
title WiSparse: Boosting LLM Inference Efficiency with Weight-Aware Mixed Activation Sparsity
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2602.14452