Reliable, Adaptable, and Attributable Language Models with Retrieval

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Asai, Akari, Zhong, Zexuan, Chen, Danqi, Koh, Pang Wei, Zettlemoyer, Luke, Hajishirzi, Hannaneh, Yih, Wen-tau
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913255333560320
author Asai, Akari
Zhong, Zexuan
Chen, Danqi
Koh, Pang Wei
Zettlemoyer, Luke
Hajishirzi, Hannaneh
Yih, Wen-tau
author_facet Asai, Akari
Zhong, Zexuan
Chen, Danqi
Koh, Pang Wei
Zettlemoyer, Luke
Hajishirzi, Hannaneh
Yih, Wen-tau
contents Parametric language models (LMs), which are trained on vast amounts of web data, exhibit remarkable flexibility and capability. However, they still face practical challenges such as hallucinations, difficulty in adapting to new data distributions, and a lack of verifiability. In this position paper, we advocate for retrieval-augmented LMs to replace parametric LMs as the next generation of LMs. By incorporating large-scale datastores during inference, retrieval-augmented LMs can be more reliable, adaptable, and attributable. Despite their potential, retrieval-augmented LMs have yet to be widely adopted due to several obstacles: specifically, current retrieval-augmented LMs struggle to leverage helpful text beyond knowledge-intensive tasks such as question answering, have limited interaction between retrieval and LM components, and lack the infrastructure for scaling. To address these, we propose a roadmap for developing general-purpose retrieval-augmented LMs. This involves a reconsideration of datastores and retrievers, the exploration of pipelines with improved retriever-LM interaction, and significant investment in infrastructure for efficient training and inference.
format Preprint
id arxiv_https___arxiv_org_abs_2403_03187
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Reliable, Adaptable, and Attributable Language Models with Retrieval
Asai, Akari
Zhong, Zexuan
Chen, Danqi
Koh, Pang Wei
Zettlemoyer, Luke
Hajishirzi, Hannaneh
Yih, Wen-tau
Computation and Language
Artificial Intelligence
Machine Learning
Parametric language models (LMs), which are trained on vast amounts of web data, exhibit remarkable flexibility and capability. However, they still face practical challenges such as hallucinations, difficulty in adapting to new data distributions, and a lack of verifiability. In this position paper, we advocate for retrieval-augmented LMs to replace parametric LMs as the next generation of LMs. By incorporating large-scale datastores during inference, retrieval-augmented LMs can be more reliable, adaptable, and attributable. Despite their potential, retrieval-augmented LMs have yet to be widely adopted due to several obstacles: specifically, current retrieval-augmented LMs struggle to leverage helpful text beyond knowledge-intensive tasks such as question answering, have limited interaction between retrieval and LM components, and lack the infrastructure for scaling. To address these, we propose a roadmap for developing general-purpose retrieval-augmented LMs. This involves a reconsideration of datastores and retrievers, the exploration of pipelines with improved retriever-LM interaction, and significant investment in infrastructure for efficient training and inference.
title Reliable, Adaptable, and Attributable Language Models with Retrieval
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2403.03187