Despite the success of large language models (LLMs) in various natural
language processing (NLP) tasks, the stored knowledge in these models may
inevitably be incomplete, out-of-date, or incorrect. This motivates the need to
utilize external knowledge to assist LLMs. Unfortunately, current methods for
incorporating external knowledge often require additional training or
fine-tuning, which can be costly and may not be feasible for LLMs. To address
this issue, we propose a novel post-processing approach, rethinking with
retrieval (RR), which retrieves relevant external knowledge based on the
decomposed reasoning steps obtained from the chain-of-thought (CoT) prompting.
This lightweight approach does not require additional training or fine-tuning
and is not limited by the input length of LLMs. We evaluate the effectiveness
of RR through extensive experiments with GPT-3 on three complex reasoning
tasks: commonsense reasoning, temporal reasoning, and tabular reasoning. Our
results show that RR can produce more faithful explanations and improve the
performance of LLMs.
Description
Rethinking with Retrieval: Faithful Large Language Model Inference
%0 Generic
%1 he2022rethinking
%A He, Hangfeng
%A Zhang, Hongming
%A Roth, Dan
%D 2022
%K llm retrieval
%T Rethinking with Retrieval: Faithful Large Language Model Inference
%U http://arxiv.org/abs/2301.00303
%X Despite the success of large language models (LLMs) in various natural
language processing (NLP) tasks, the stored knowledge in these models may
inevitably be incomplete, out-of-date, or incorrect. This motivates the need to
utilize external knowledge to assist LLMs. Unfortunately, current methods for
incorporating external knowledge often require additional training or
fine-tuning, which can be costly and may not be feasible for LLMs. To address
this issue, we propose a novel post-processing approach, rethinking with
retrieval (RR), which retrieves relevant external knowledge based on the
decomposed reasoning steps obtained from the chain-of-thought (CoT) prompting.
This lightweight approach does not require additional training or fine-tuning
and is not limited by the input length of LLMs. We evaluate the effectiveness
of RR through extensive experiments with GPT-3 on three complex reasoning
tasks: commonsense reasoning, temporal reasoning, and tabular reasoning. Our
results show that RR can produce more faithful explanations and improve the
performance of LLMs.
@misc{he2022rethinking,
abstract = {Despite the success of large language models (LLMs) in various natural
language processing (NLP) tasks, the stored knowledge in these models may
inevitably be incomplete, out-of-date, or incorrect. This motivates the need to
utilize external knowledge to assist LLMs. Unfortunately, current methods for
incorporating external knowledge often require additional training or
fine-tuning, which can be costly and may not be feasible for LLMs. To address
this issue, we propose a novel post-processing approach, rethinking with
retrieval (RR), which retrieves relevant external knowledge based on the
decomposed reasoning steps obtained from the chain-of-thought (CoT) prompting.
This lightweight approach does not require additional training or fine-tuning
and is not limited by the input length of LLMs. We evaluate the effectiveness
of RR through extensive experiments with GPT-3 on three complex reasoning
tasks: commonsense reasoning, temporal reasoning, and tabular reasoning. Our
results show that RR can produce more faithful explanations and improve the
performance of LLMs.},
added-at = {2023-08-17T15:00:13.000+0200},
author = {He, Hangfeng and Zhang, Hongming and Roth, Dan},
biburl = {https://www.bibsonomy.org/bibtex/228dc13cb9071cc2b9c07893298939fe3/lisa-ee},
description = {Rethinking with Retrieval: Faithful Large Language Model Inference},
interhash = {9b3be74c72df8a6b0771efb41be3e894},
intrahash = {28dc13cb9071cc2b9c07893298939fe3},
keywords = {llm retrieval},
note = {cite arxiv:2301.00303},
timestamp = {2023-08-17T15:00:13.000+0200},
title = {Rethinking with Retrieval: Faithful Large Language Model Inference},
url = {http://arxiv.org/abs/2301.00303},
year = 2022
}