@inproceedings{cai-etal-2024-ceret,
title = "{CERET}: Cost-Effective Extrinsic Refinement for Text Generation",
author = "Cai, Jason and
Su, Hang and
Sunkara, Monica and
Shalyminov, Igor and
Mansour, Saab",
editor = "Duh, Kevin and
Gomez, Helena and
Bethard, Steven",
booktitle = "Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = jun,
year = "2024",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.naacl-long.409",
doi = "10.18653/v1/2024.naacl-long.409",
pages = "7377--7390",
abstract = "Large Language Models (LLMs) are powerful models for generation tasks, but they may not generate good quality outputs in their first attempt. Apart from model fine-tuning, existing approaches to improve prediction accuracy and quality typically involve LLM self-improvement / self-reflection that incorporate feedback from models themselves. Despite their effectiveness, these methods are hindered by their high computational cost and lack of scalability. In this work, we propose CERET, a method for refining text generations by considering semantic stability, entailment and inter-sample uncertainty measures. Experimental results show that CERET outperforms Self-consistency and Self-rerank baselines consistently under various task setups, by 1.6{\%} in Rouge-1 for abstractive summarization and 3.5{\%} in hit rate for question answering. Compared to LLM Self-rerank method, our approach only requires 9.4{\%} of its latency and is more cost-effective.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cai-etal-2024-ceret">
<titleInfo>
<title>CERET: Cost-Effective Extrinsic Refinement for Text Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jason</namePart>
<namePart type="family">Cai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hang</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Monica</namePart>
<namePart type="family">Sunkara</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Igor</namePart>
<namePart type="family">Shalyminov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Saab</namePart>
<namePart type="family">Mansour</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-06</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Kevin</namePart>
<namePart type="family">Duh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Helena</namePart>
<namePart type="family">Gomez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Steven</namePart>
<namePart type="family">Bethard</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Mexico City, Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large Language Models (LLMs) are powerful models for generation tasks, but they may not generate good quality outputs in their first attempt. Apart from model fine-tuning, existing approaches to improve prediction accuracy and quality typically involve LLM self-improvement / self-reflection that incorporate feedback from models themselves. Despite their effectiveness, these methods are hindered by their high computational cost and lack of scalability. In this work, we propose CERET, a method for refining text generations by considering semantic stability, entailment and inter-sample uncertainty measures. Experimental results show that CERET outperforms Self-consistency and Self-rerank baselines consistently under various task setups, by 1.6% in Rouge-1 for abstractive summarization and 3.5% in hit rate for question answering. Compared to LLM Self-rerank method, our approach only requires 9.4% of its latency and is more cost-effective.</abstract>
<identifier type="citekey">cai-etal-2024-ceret</identifier>
<identifier type="doi">10.18653/v1/2024.naacl-long.409</identifier>
<location>
<url>https://aclanthology.org/2024.naacl-long.409</url>
</location>
<part>
<date>2024-06</date>
<extent unit="page">
<start>7377</start>
<end>7390</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CERET: Cost-Effective Extrinsic Refinement for Text Generation
%A Cai, Jason
%A Su, Hang
%A Sunkara, Monica
%A Shalyminov, Igor
%A Mansour, Saab
%Y Duh, Kevin
%Y Gomez, Helena
%Y Bethard, Steven
%S Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2024
%8 June
%I Association for Computational Linguistics
%C Mexico City, Mexico
%F cai-etal-2024-ceret
%X Large Language Models (LLMs) are powerful models for generation tasks, but they may not generate good quality outputs in their first attempt. Apart from model fine-tuning, existing approaches to improve prediction accuracy and quality typically involve LLM self-improvement / self-reflection that incorporate feedback from models themselves. Despite their effectiveness, these methods are hindered by their high computational cost and lack of scalability. In this work, we propose CERET, a method for refining text generations by considering semantic stability, entailment and inter-sample uncertainty measures. Experimental results show that CERET outperforms Self-consistency and Self-rerank baselines consistently under various task setups, by 1.6% in Rouge-1 for abstractive summarization and 3.5% in hit rate for question answering. Compared to LLM Self-rerank method, our approach only requires 9.4% of its latency and is more cost-effective.
%R 10.18653/v1/2024.naacl-long.409
%U https://aclanthology.org/2024.naacl-long.409
%U https://doi.org/10.18653/v1/2024.naacl-long.409
%P 7377-7390
Markdown (Informal)
[CERET: Cost-Effective Extrinsic Refinement for Text Generation](https://aclanthology.org/2024.naacl-long.409) (Cai et al., NAACL 2024)
ACL
- Jason Cai, Hang Su, Monica Sunkara, Igor Shalyminov, and Saab Mansour. 2024. CERET: Cost-Effective Extrinsic Refinement for Text Generation. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 7377–7390, Mexico City, Mexico. Association for Computational Linguistics.