UniICL: An Efficient Unified Framework Unifying Compression, Selection, and Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gao, Jun, Lv, Qi, Wang, Zili, Wu, Tianxiang, Cao, Ziqiang, Li, Wenjie
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913858569895936
author Gao, Jun
Lv, Qi
Wang, Zili
Wu, Tianxiang
Cao, Ziqiang
Li, Wenjie
author_facet Gao, Jun
Lv, Qi
Wang, Zili
Wu, Tianxiang
Cao, Ziqiang
Li, Wenjie
contents In-context learning (ICL) enhances the reasoning abilities of Large Language Models (LLMs) by prepending a few demonstrations. It motivates researchers to introduce more examples to provide additional contextual information for the generation. However, existing methods show a significant limitation due to the problem of excessive growth in context length, which causes a large hardware burden. In addition, shallow-relevant examples selected by off-the-shelf tools hinder LLMs from capturing useful contextual information for generation. In this paper, we propose \textbf{UniICL}, a novel \textbf{Uni}fied \textbf{ICL} framework that unifies demonstration compression, demonstration selection, and final response generation. Furthermore, to boost inference efficiency, we design a tailored compression strategy that allows UniICL to cache compression results into \textbf{Demonstration Bank} (\textbf{DB}), which avoids repeated compression of the same demonstration. Extensive out-of-domain evaluations prove the advantages of UniICL in both effectiveness and efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2405_17062
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle UniICL: An Efficient Unified Framework Unifying Compression, Selection, and Generation
Gao, Jun
Lv, Qi
Wang, Zili
Wu, Tianxiang
Cao, Ziqiang
Li, Wenjie
Computation and Language
In-context learning (ICL) enhances the reasoning abilities of Large Language Models (LLMs) by prepending a few demonstrations. It motivates researchers to introduce more examples to provide additional contextual information for the generation. However, existing methods show a significant limitation due to the problem of excessive growth in context length, which causes a large hardware burden. In addition, shallow-relevant examples selected by off-the-shelf tools hinder LLMs from capturing useful contextual information for generation. In this paper, we propose \textbf{UniICL}, a novel \textbf{Uni}fied \textbf{ICL} framework that unifies demonstration compression, demonstration selection, and final response generation. Furthermore, to boost inference efficiency, we design a tailored compression strategy that allows UniICL to cache compression results into \textbf{Demonstration Bank} (\textbf{DB}), which avoids repeated compression of the same demonstration. Extensive out-of-domain evaluations prove the advantages of UniICL in both effectiveness and efficiency.
title UniICL: An Efficient Unified Framework Unifying Compression, Selection, and Generation
topic Computation and Language
url https://arxiv.org/abs/2405.17062