@inproceedings{yuan-etal-2026-e3,
title = "{E}$^3$-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models",
author = "Yuan, Tao and
Bai, Haoli and
Yinfei, Pan and
Cao, Xuyang and
Zhang, Tianyu and
Hou, Lu and
Hu, Ting and
Yu, Xianzhi",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.38/",
pages = "830--846",
ISBN = "979-8-89176-459-0",
abstract = "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 E$^3$-Pruner, a task-Effective, training-Economical and inference-Efficient layer pruning framework. E$^3$-Pruner 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, E$^3$-Pruner 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)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yuan-etal-2026-e3">
<titleInfo>
<title>E³-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Tao</namePart>
<namePart type="family">Yuan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haoli</namePart>
<namePart type="family">Bai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pan</namePart>
<namePart type="family">Yinfei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuyang</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tianyu</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Hou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ting</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xianzhi</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>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 E³-Pruner, a task-Effective, training-Economical and inference-Efficient layer pruning framework. E³-Pruner 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, E³-Pruner 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).</abstract>
<identifier type="citekey">yuan-etal-2026-e3</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.38/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>830</start>
<end>846</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T E³-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models
%A Yuan, Tao
%A Bai, Haoli
%A Yinfei, Pan
%A Cao, Xuyang
%A Zhang, Tianyu
%A Hou, Lu
%A Hu, Ting
%A Yu, Xianzhi
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F yuan-etal-2026-e3
%X 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 E³-Pruner, a task-Effective, training-Economical and inference-Efficient layer pruning framework. E³-Pruner 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, E³-Pruner 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).
%U https://aclanthology.org/2026.emnlp-main.38/
%P 830-846
Markdown (Informal)
[E3-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models](https://aclanthology.org/2026.emnlp-main.38/) (Yuan et al., EMNLP 2026)
ACL
- Tao Yuan, Haoli Bai, Pan Yinfei, Xuyang Cao, Tianyu Zhang, Lu Hou, Ting Hu, and Xianzhi Yu. 2026. E3-Pruner: Towards Efficient, Economical, and Effective Layer Pruning for Large Language Models. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 830–846, Budapest, Hungary. Association for Computational Linguistics.