DAC: A Dynamic Attention-aware Approach for Task-Agnostic Prompt Compression

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhao, Yi, Li, Zuchao, Zhao, Hai, Qi, Baoyuan, Liu, Guoming
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911059422478336
author Zhao, Yi
Li, Zuchao
Zhao, Hai
Qi, Baoyuan
Liu, Guoming
author_facet Zhao, Yi
Li, Zuchao
Zhao, Hai
Qi, Baoyuan
Liu, Guoming
contents Task-agnostic prompt compression leverages the redundancy in natural language to reduce computational overhead and enhance information density within prompts, especially in long-context scenarios. Existing methods predominantly rely on information entropy as the metric to compress lexical units, aiming to achieve minimal information loss. However, these approaches overlook two critical aspects: (i) the importance of attention-critical tokens at the algorithmic level, and (ii) shifts in information entropy during the compression process. Motivated by these challenges, we propose a dynamic attention-aware approach for task-agnostic prompt compression (DAC). This approach effectively integrates entropy and attention information, dynamically sensing entropy shifts during compression to achieve fine-grained prompt compression. Extensive experiments across various domains, including LongBench, GSM8K, and BBH, show that DAC consistently yields robust and substantial improvements across a diverse range of tasks and LLMs, offering compelling evidence of its efficacy.
format Preprint
id arxiv_https___arxiv_org_abs_2507_11942
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DAC: A Dynamic Attention-aware Approach for Task-Agnostic Prompt Compression
Zhao, Yi
Li, Zuchao
Zhao, Hai
Qi, Baoyuan
Liu, Guoming
Computation and Language
Task-agnostic prompt compression leverages the redundancy in natural language to reduce computational overhead and enhance information density within prompts, especially in long-context scenarios. Existing methods predominantly rely on information entropy as the metric to compress lexical units, aiming to achieve minimal information loss. However, these approaches overlook two critical aspects: (i) the importance of attention-critical tokens at the algorithmic level, and (ii) shifts in information entropy during the compression process. Motivated by these challenges, we propose a dynamic attention-aware approach for task-agnostic prompt compression (DAC). This approach effectively integrates entropy and attention information, dynamically sensing entropy shifts during compression to achieve fine-grained prompt compression. Extensive experiments across various domains, including LongBench, GSM8K, and BBH, show that DAC consistently yields robust and substantial improvements across a diverse range of tasks and LLMs, offering compelling evidence of its efficacy.
title DAC: A Dynamic Attention-aware Approach for Task-Agnostic Prompt Compression
topic Computation and Language
url https://arxiv.org/abs/2507.11942