Saved in:
Bibliographic Details
Main Authors: Xu, Zhentao, Li, Fengyi, Chen, Albert, Wang, Xiaofeng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.02053
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912634281918464
author Xu, Zhentao
Li, Fengyi
Chen, Albert
Wang, Xiaofeng
author_facet Xu, Zhentao
Li, Fengyi
Chen, Albert
Wang, Xiaofeng
contents In large-scale industrial LLM systems, prompt templates often expand to thousands of tokens as teams iteratively incorporate sections such as task instructions, few-shot examples, and heuristic rules to enhance robustness and coverage. This expansion leads to bloated prompts that are difficult to maintain and incur significant inference latency and serving costs. To address this, we introduce Prompt Compression via Attribution Estimation (ProCut), a flexible, LLM-agnostic, training-free framework that compresses prompts through attribution analysis. ProCut segments prompt templates into semantically meaningful units, quantifies their impact on task performance, and prunes low-utility components. Through extensive experiments on five public benchmark datasets and real-world industrial prompts, we show that ProCut achieves substantial prompt size reductions (78% fewer tokens in production) while maintaining or even slightly improving task performance (up to 62% better than alternative methods). We further introduce an LLM-driven attribution estimator that reduces compression latency by over 50%, and demonstrate that ProCut integrates seamlessly with existing prompt-optimization frameworks to produce concise, high-performing prompts.
format Preprint
id arxiv_https___arxiv_org_abs_2508_02053
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ProCut: LLM Prompt Compression via Attribution Estimation
Xu, Zhentao
Li, Fengyi
Chen, Albert
Wang, Xiaofeng
Computation and Language
Machine Learning
In large-scale industrial LLM systems, prompt templates often expand to thousands of tokens as teams iteratively incorporate sections such as task instructions, few-shot examples, and heuristic rules to enhance robustness and coverage. This expansion leads to bloated prompts that are difficult to maintain and incur significant inference latency and serving costs. To address this, we introduce Prompt Compression via Attribution Estimation (ProCut), a flexible, LLM-agnostic, training-free framework that compresses prompts through attribution analysis. ProCut segments prompt templates into semantically meaningful units, quantifies their impact on task performance, and prunes low-utility components. Through extensive experiments on five public benchmark datasets and real-world industrial prompts, we show that ProCut achieves substantial prompt size reductions (78% fewer tokens in production) while maintaining or even slightly improving task performance (up to 62% better than alternative methods). We further introduce an LLM-driven attribution estimator that reduces compression latency by over 50%, and demonstrate that ProCut integrates seamlessly with existing prompt-optimization frameworks to produce concise, high-performing prompts.
title ProCut: LLM Prompt Compression via Attribution Estimation
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2508.02053