ReFusion: Improving Natural Language Understanding with Computation-Efficient Retrieval Representation Fusion

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wu, Shangyu, Xiong, Ying, Cui, Yufei, Liu, Xue, Tang, Buzhou, Kuo, Tei-Wei, Xue, Chun Jason
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866914812363014144
author Wu, Shangyu
Xiong, Ying
Cui, Yufei
Liu, Xue
Tang, Buzhou
Kuo, Tei-Wei
Xue, Chun Jason
author_facet Wu, Shangyu
Xiong, Ying
Cui, Yufei
Liu, Xue
Tang, Buzhou
Kuo, Tei-Wei
Xue, Chun Jason
contents Retrieval-based augmentations (RA) incorporating knowledge from an external database into language models have greatly succeeded in various knowledge-intensive (KI) tasks. However, integrating retrievals in non-knowledge-intensive (NKI) tasks is still challenging. Existing works focus on concatenating retrievals with inputs to improve model performance. Unfortunately, the use of retrieval concatenation-based augmentations causes an increase in the input length, substantially raising the computational demands of attention mechanisms. This paper proposes a new paradigm of RA named \textbf{ReFusion}, a computation-efficient Retrieval representation Fusion with bi-level optimization. Unlike previous works, ReFusion directly fuses the retrieval representations into the hidden states of models. Specifically, ReFusion leverages an adaptive retrieval integrator to seek the optimal combination of the proposed ranking schemes across different model layers. Experimental results demonstrate that the proposed ReFusion can achieve superior and robust performance in various NKI tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2401_02993
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle ReFusion: Improving Natural Language Understanding with Computation-Efficient Retrieval Representation Fusion
Wu, Shangyu
Xiong, Ying
Cui, Yufei
Liu, Xue
Tang, Buzhou
Kuo, Tei-Wei
Xue, Chun Jason
Computation and Language
Artificial Intelligence
Retrieval-based augmentations (RA) incorporating knowledge from an external database into language models have greatly succeeded in various knowledge-intensive (KI) tasks. However, integrating retrievals in non-knowledge-intensive (NKI) tasks is still challenging. Existing works focus on concatenating retrievals with inputs to improve model performance. Unfortunately, the use of retrieval concatenation-based augmentations causes an increase in the input length, substantially raising the computational demands of attention mechanisms. This paper proposes a new paradigm of RA named \textbf{ReFusion}, a computation-efficient Retrieval representation Fusion with bi-level optimization. Unlike previous works, ReFusion directly fuses the retrieval representations into the hidden states of models. Specifically, ReFusion leverages an adaptive retrieval integrator to seek the optimal combination of the proposed ranking schemes across different model layers. Experimental results demonstrate that the proposed ReFusion can achieve superior and robust performance in various NKI tasks.
title ReFusion: Improving Natural Language Understanding with Computation-Efficient Retrieval Representation Fusion
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2401.02993