English

Planning Ahead in Generative Retrieval: Guiding Autoregressive Generation through Simultaneous Decoding

Information Retrieval 2024-04-24 v1 Computation and Language

Abstract

This paper introduces PAG-a novel optimization and decoding approach that guides autoregressive generation of document identifiers in generative retrieval models through simultaneous decoding. To this aim, PAG constructs a set-based and sequential identifier for each document. Motivated by the bag-of-words assumption in information retrieval, the set-based identifier is built on lexical tokens. The sequential identifier, on the other hand, is obtained via quantizing relevance-based representations of documents. Extensive experiments on MSMARCO and TREC Deep Learning Track data reveal that PAG outperforms the state-of-the-art generative retrieval model by a large margin (e.g., 15.6% MRR improvements on MS MARCO), while achieving 22x speed up in terms of query latency.

Keywords

Cite

@article{arxiv.2404.14600,
  title  = {Planning Ahead in Generative Retrieval: Guiding Autoregressive Generation through Simultaneous Decoding},
  author = {Hansi Zeng and Chen Luo and Hamed Zamani},
  journal= {arXiv preprint arXiv:2404.14600},
  year   = {2024}
}

Comments

Accepted to SIGIR 2024

R2 v1 2026-06-28T16:02:57.048Z