DAST: Context-Aware Compression in LLMs via Dynamic Allocation of Soft Tokens

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Shaoshen, Li, Yangning, Xu, Zishan, Li, Yinghui, Su, Xin, Shan, Zifei, Zheng, Hai-tao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909496701353984
author Chen, Shaoshen
Li, Yangning
Xu, Zishan
Li, Yinghui
Su, Xin
Shan, Zifei
Zheng, Hai-tao
author_facet Chen, Shaoshen
Li, Yangning
Xu, Zishan
Li, Yinghui
Su, Xin
Shan, Zifei
Zheng, Hai-tao
contents Large Language Models (LLMs) face computational inefficiencies and redundant processing when handling long context inputs, prompting a focus on compression techniques. While existing semantic vector-based compression methods achieve promising performance, these methods fail to account for the intrinsic information density variations between context chunks, instead allocating soft tokens uniformly across context chunks. This uniform distribution inevitably diminishes allocation to information-critical regions. To address this, we propose Dynamic Allocation of Soft Tokens (DAST), a simple yet effective method that leverages the LLM's intrinsic understanding of contextual relevance to guide compression. DAST combines perplexity-based local information with attention-driven global information to dynamically allocate soft tokens to the informative-rich chunks, enabling effective, context-aware compression. Experimental results across multiple benchmarks demonstrate that DAST surpasses state-of-the-art methods.
format Preprint
id arxiv_https___arxiv_org_abs_2502_11493
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DAST: Context-Aware Compression in LLMs via Dynamic Allocation of Soft Tokens
Chen, Shaoshen
Li, Yangning
Xu, Zishan
Li, Yinghui
Su, Xin
Shan, Zifei
Zheng, Hai-tao
Computation and Language
Large Language Models (LLMs) face computational inefficiencies and redundant processing when handling long context inputs, prompting a focus on compression techniques. While existing semantic vector-based compression methods achieve promising performance, these methods fail to account for the intrinsic information density variations between context chunks, instead allocating soft tokens uniformly across context chunks. This uniform distribution inevitably diminishes allocation to information-critical regions. To address this, we propose Dynamic Allocation of Soft Tokens (DAST), a simple yet effective method that leverages the LLM's intrinsic understanding of contextual relevance to guide compression. DAST combines perplexity-based local information with attention-driven global information to dynamically allocate soft tokens to the informative-rich chunks, enabling effective, context-aware compression. Experimental results across multiple benchmarks demonstrate that DAST surpasses state-of-the-art methods.
title DAST: Context-Aware Compression in LLMs via Dynamic Allocation of Soft Tokens
topic Computation and Language
url https://arxiv.org/abs/2502.11493