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

TL;DR

DAST introduces a context-aware compression method for LLMs that dynamically allocates soft tokens based on local and global information, improving efficiency and performance on various benchmarks.

Contribution

The paper presents DAST, a novel method that adaptively allocates soft tokens in LLMs using relevance signals, addressing the limitations of uniform distribution in prior approaches.

Findings

DAST outperforms existing compression methods on multiple benchmarks.

Dynamic token allocation improves model efficiency and relevance.

DAST effectively identifies and preserves critical contextual information.

Abstract

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.