From Reading to Compressing: Exploring the Multi-document Reader for Prompt Compression

TL;DR

This paper introduces R2C, a prompt compression method using FiD architecture that preserves essential information, reduces prompt length by 80%, and improves LLM performance by 6% in out-of-domain tasks.

Contribution

The paper proposes a novel prompt compression technique, R2C, which effectively captures global context without pseudo-labels, enhancing LLM efficiency and performance.

Findings

Retains key context while reducing prompt length by 80%.

Improves out-of-domain LLM performance by 6%.

Uses cross-attention scores for identifying important prompt segments.

Abstract

Large language models (LLMs) have achieved significant performance gains using advanced prompting techniques over various tasks. However, the increasing length of prompts leads to high computational costs and often obscures crucial information. Prompt compression has been proposed to alleviate these issues, but it faces challenges in (i) capturing the global context and (ii) training the compressor effectively. To tackle these challenges, we introduce a novel prompt compression method, namely Reading To Compressing (R2C), utilizing the Fusion-in-Decoder (FiD) architecture to identify the important information in the prompt. Specifically, the cross-attention scores of the FiD are used to discern essential chunks and sentences from the prompt. R2C effectively captures the global context without compromising semantic consistency while detouring the necessity of pseudo-labels for training the compressor. Empirical results show that R2C retains key contexts, enhancing the LLM performance by 6% in out-of-domain evaluations while reducing the prompt length by 80%.