Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Tang, Jiwei, Xu, Jin, Lu, Tingwei, Zhang, Zhicheng, Zhao, Yiming, Hai, Lin, Zheng, Hai-Tao
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2409.19272
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915142588956672
author Tang, Jiwei
Xu, Jin
Lu, Tingwei
Zhang, Zhicheng
Zhao, Yiming
Hai, Lin
Zheng, Hai-Tao
author_facet Tang, Jiwei
Xu, Jin
Lu, Tingwei
Zhang, Zhicheng
Zhao, Yiming
Hai, Lin
Zheng, Hai-Tao
contents Large language models (LLMs) demonstrate exceptional capabilities in various scenarios. However, they suffer from much redundant information and are sensitive to the position of key information in long context scenarios. To address these challenges, we present Perception Compressor, a training-free prompt compression framework. It includes a perception retriever that leverages guiding questions and instruction to retrieve the most relevant demonstrations, a dual-slope ratio allocator to dynamically allocate compression ratios and open-book ratios, and a semi-guided iterative compression that retains key information at the token level while removing tokens that distract the LLM. We conduct extensive experiments on long context benchmarks, i.e., NaturalQuestions, LongBench, and MuSiQue. Experiment results show that Perception Compressor outperforms existing methods by a large margin, achieving state-of-the-art performance.
format Preprint
id arxiv_https___arxiv_org_abs_2409_19272
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Perception Compressor: A Training-Free Prompt Compression Framework in Long Context Scenarios
Tang, Jiwei
Xu, Jin
Lu, Tingwei
Zhang, Zhicheng
Zhao, Yiming
Hai, Lin
Zheng, Hai-Tao
Computation and Language
Large language models (LLMs) demonstrate exceptional capabilities in various scenarios. However, they suffer from much redundant information and are sensitive to the position of key information in long context scenarios. To address these challenges, we present Perception Compressor, a training-free prompt compression framework. It includes a perception retriever that leverages guiding questions and instruction to retrieve the most relevant demonstrations, a dual-slope ratio allocator to dynamically allocate compression ratios and open-book ratios, and a semi-guided iterative compression that retains key information at the token level while removing tokens that distract the LLM. We conduct extensive experiments on long context benchmarks, i.e., NaturalQuestions, LongBench, and MuSiQue. Experiment results show that Perception Compressor outperforms existing methods by a large margin, achieving state-of-the-art performance.
title Perception Compressor: A Training-Free Prompt Compression Framework in Long Context Scenarios
topic Computation and Language
url https://arxiv.org/abs/2409.19272