E$^3$-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yuan, Tao, Bai, Haoli, Pan, Yinfei, Cao, Xuyang, Zhang, Tianyu, Hou, Lu, Hu, Ting, Yu, Xianzhi
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918213293441024
author Yuan, Tao
Bai, Haoli
Pan, Yinfei
Cao, Xuyang
Zhang, Tianyu
Hou, Lu
Hu, Ting
Yu, Xianzhi
author_facet Yuan, Tao
Bai, Haoli
Pan, Yinfei
Cao, Xuyang
Zhang, Tianyu
Hou, Lu
Hu, Ting
Yu, Xianzhi
contents With the increasing size of large language models, layer pruning has gained increased attention as a hardware-friendly approach for model compression. However, existing layer pruning methods struggle to simultaneously address key practical deployment challenges, including performance degradation, high training costs, and limited acceleration. To overcome these limitations, we propose \name, a task-\underline{E}ffective, training-\underline{E}conomical and inference-\underline{E}fficient layer pruning framework. \namespace introduces two key innovations: (1) a differentiable mask optimization method using a Gumbel-TopK sampler, enabling efficient and precise pruning mask search; and (2) an entropy-aware adaptive knowledge distillation strategy that enhances task performance. Extensive experiments over diverse model architectures and benchmarks demonstrate the superiority of our method over state-of-the-art approaches. Notably, \namespace achieves 96\% accuracy, a mere 0.8\% drop from the original model (96.8\%) on MATH-500 when pruning 25\% layers of Qwen3-32B, outperforming existing SOTA (95\%), with a 1.33$\times$ inference speedup by consuming merely 0.5B tokens (0.5\% of the post-training data volume).
format Preprint
id arxiv_https___arxiv_org_abs_2511_17205
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle E$^3$-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models
Yuan, Tao
Bai, Haoli
Pan, Yinfei
Cao, Xuyang
Zhang, Tianyu
Hou, Lu
Hu, Ting
Yu, Xianzhi
Computation and Language
With the increasing size of large language models, layer pruning has gained increased attention as a hardware-friendly approach for model compression. However, existing layer pruning methods struggle to simultaneously address key practical deployment challenges, including performance degradation, high training costs, and limited acceleration. To overcome these limitations, we propose \name, a task-\underline{E}ffective, training-\underline{E}conomical and inference-\underline{E}fficient layer pruning framework. \namespace introduces two key innovations: (1) a differentiable mask optimization method using a Gumbel-TopK sampler, enabling efficient and precise pruning mask search; and (2) an entropy-aware adaptive knowledge distillation strategy that enhances task performance. Extensive experiments over diverse model architectures and benchmarks demonstrate the superiority of our method over state-of-the-art approaches. Notably, \namespace achieves 96\% accuracy, a mere 0.8\% drop from the original model (96.8\%) on MATH-500 when pruning 25\% layers of Qwen3-32B, outperforming existing SOTA (95\%), with a 1.33$\times$ inference speedup by consuming merely 0.5B tokens (0.5\% of the post-training data volume).
title E$^3$-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models
topic Computation and Language
url https://arxiv.org/abs/2511.17205