@inproceedings{tang-etal-2024-bootstrapped,
title = "Bootstrapped Pre-training with Dynamic Identifier Prediction for Generative Retrieval",
author = "Tang, Yubao and
Zhang, Ruqing and
Guo, Jiafeng and
de Rijke, Maarten and
Fan, Yixing and
Cheng, Xueqi",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.614/",
doi = "10.18653/v1/2024.findings-acl.614",
pages = "10303--10317",
abstract = "Generative retrieval uses differentiable search indexes to directly generate relevant document identifiers in response to a query. Recent studies have highlighted the potential of a strong generative retrieval model, trained with carefully crafted pre-training tasks, to enhance downstream retrieval tasks via fine-tuning. However, the full power of pre-training for generative retrieval remains underexploited due to its reliance on pre-defined static document identifiers, which may not align with evolving model parameters. In this work, we introduce BootRet, a bootstrapped pre-training method for generative retrieval that dynamically adjusts document identifiers during pre-training to accommodate the continuing memorization of the corpus. BootRet involves three key training phases: (i) initial identifier generation, (ii) pre-training via corpus indexing and relevance prediction tasks, and (iii) bootstrapping for identifier updates. To facilitate the pre-training phase, we further introduce noisy documents and pseudo-queries, generated by large language models, to resemble semantic connections in both indexing and retrieval tasks. Experimental results demonstrate that BootRet significantly outperforms existing pre-training generative retrieval baselines and performs well even in zero-shot settings."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tang-etal-2024-bootstrapped">
<titleInfo>
<title>Bootstrapped Pre-training with Dynamic Identifier Prediction for Generative Retrieval</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yubao</namePart>
<namePart type="family">Tang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruqing</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiafeng</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maarten</namePart>
<namePart type="family">de Rijke</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixing</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xueqi</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Generative retrieval uses differentiable search indexes to directly generate relevant document identifiers in response to a query. Recent studies have highlighted the potential of a strong generative retrieval model, trained with carefully crafted pre-training tasks, to enhance downstream retrieval tasks via fine-tuning. However, the full power of pre-training for generative retrieval remains underexploited due to its reliance on pre-defined static document identifiers, which may not align with evolving model parameters. In this work, we introduce BootRet, a bootstrapped pre-training method for generative retrieval that dynamically adjusts document identifiers during pre-training to accommodate the continuing memorization of the corpus. BootRet involves three key training phases: (i) initial identifier generation, (ii) pre-training via corpus indexing and relevance prediction tasks, and (iii) bootstrapping for identifier updates. To facilitate the pre-training phase, we further introduce noisy documents and pseudo-queries, generated by large language models, to resemble semantic connections in both indexing and retrieval tasks. Experimental results demonstrate that BootRet significantly outperforms existing pre-training generative retrieval baselines and performs well even in zero-shot settings.</abstract>
<identifier type="citekey">tang-etal-2024-bootstrapped</identifier>
<identifier type="doi">10.18653/v1/2024.findings-acl.614</identifier>
<location>
<url>https://aclanthology.org/2024.findings-acl.614/</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>10303</start>
<end>10317</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Bootstrapped Pre-training with Dynamic Identifier Prediction for Generative Retrieval
%A Tang, Yubao
%A Zhang, Ruqing
%A Guo, Jiafeng
%A de Rijke, Maarten
%A Fan, Yixing
%A Cheng, Xueqi
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Findings of the Association for Computational Linguistics: ACL 2024
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F tang-etal-2024-bootstrapped
%X Generative retrieval uses differentiable search indexes to directly generate relevant document identifiers in response to a query. Recent studies have highlighted the potential of a strong generative retrieval model, trained with carefully crafted pre-training tasks, to enhance downstream retrieval tasks via fine-tuning. However, the full power of pre-training for generative retrieval remains underexploited due to its reliance on pre-defined static document identifiers, which may not align with evolving model parameters. In this work, we introduce BootRet, a bootstrapped pre-training method for generative retrieval that dynamically adjusts document identifiers during pre-training to accommodate the continuing memorization of the corpus. BootRet involves three key training phases: (i) initial identifier generation, (ii) pre-training via corpus indexing and relevance prediction tasks, and (iii) bootstrapping for identifier updates. To facilitate the pre-training phase, we further introduce noisy documents and pseudo-queries, generated by large language models, to resemble semantic connections in both indexing and retrieval tasks. Experimental results demonstrate that BootRet significantly outperforms existing pre-training generative retrieval baselines and performs well even in zero-shot settings.
%R 10.18653/v1/2024.findings-acl.614
%U https://aclanthology.org/2024.findings-acl.614/
%U https://doi.org/10.18653/v1/2024.findings-acl.614
%P 10303-10317
Markdown (Informal)
[Bootstrapped Pre-training with Dynamic Identifier Prediction for Generative Retrieval](https://aclanthology.org/2024.findings-acl.614/) (Tang et al., Findings 2024)
ACL