Saved in:
Bibliographic Details
Main Authors: Cai, Hongru, Li, Yongqi, Yuan, Ruifeng, Wang, Wenjie, Zhang, Zhen, Li, Wenjie, Chua, Tat-Seng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.18941
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916783595716608
author Cai, Hongru
Li, Yongqi
Yuan, Ruifeng
Wang, Wenjie
Zhang, Zhen
Li, Wenjie
Chua, Tat-Seng
author_facet Cai, Hongru
Li, Yongqi
Yuan, Ruifeng
Wang, Wenjie
Zhang, Zhen
Li, Wenjie
Chua, Tat-Seng
contents Generative retrieval reformulates retrieval as an autoregressive generation task, where large language models (LLMs) generate target documents directly from a query. As a novel paradigm, the mechanisms that underpin its performance and scalability remain largely unexplored. We systematically investigate training and inference scaling laws in generative retrieval, exploring how model size, training data scale, and inference-time compute jointly influence performance. We propose a novel evaluation metric inspired by contrastive entropy and generation loss, providing a continuous performance signal that enables robust comparisons across diverse generative retrieval methods. Our experiments show that n-gram-based methods align strongly with training and inference scaling laws. We find that increasing model size, training data scale, and inference-time compute all contribute to improved performance, highlighting the complementary roles of these factors in enhancing generative retrieval. Across these settings, LLaMA models consistently outperform T5 models, suggesting a particular advantage for larger decoder-only models in generative retrieval. Our findings underscore that model sizes, data availability, and inference computation interact to unlock the full potential of generative retrieval, offering new insights for designing and optimizing future systems.
format Preprint
id arxiv_https___arxiv_org_abs_2503_18941
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Exploring Training and Inference Scaling Laws in Generative Retrieval
Cai, Hongru
Li, Yongqi
Yuan, Ruifeng
Wang, Wenjie
Zhang, Zhen
Li, Wenjie
Chua, Tat-Seng
Information Retrieval
Computation and Language
Generative retrieval reformulates retrieval as an autoregressive generation task, where large language models (LLMs) generate target documents directly from a query. As a novel paradigm, the mechanisms that underpin its performance and scalability remain largely unexplored. We systematically investigate training and inference scaling laws in generative retrieval, exploring how model size, training data scale, and inference-time compute jointly influence performance. We propose a novel evaluation metric inspired by contrastive entropy and generation loss, providing a continuous performance signal that enables robust comparisons across diverse generative retrieval methods. Our experiments show that n-gram-based methods align strongly with training and inference scaling laws. We find that increasing model size, training data scale, and inference-time compute all contribute to improved performance, highlighting the complementary roles of these factors in enhancing generative retrieval. Across these settings, LLaMA models consistently outperform T5 models, suggesting a particular advantage for larger decoder-only models in generative retrieval. Our findings underscore that model sizes, data availability, and inference computation interact to unlock the full potential of generative retrieval, offering new insights for designing and optimizing future systems.
title Exploring Training and Inference Scaling Laws in Generative Retrieval
topic Information Retrieval
Computation and Language
url https://arxiv.org/abs/2503.18941