Does Generative Retrieval Overcome The Limitations Of Dense Retrieval?

Abstract

Generative retrieval (GR) has emerged as a new paradigm in neural information retrieval, offering an alternative to dense retrieval (DR) by directly generating identifiers of relevant documents. In this paper, we theoretically and empirically investigate how GR fundamentally diverges from DR in both learning objectives and representational capacity. GR performs globally normalized maximum-likelihood optimization and encodes corpus and relevance information directly in the model parameters, whereas DR adopts locally normalized objectives and represents the corpus with external embeddings before computing similarity via a bilinear interaction. Our analysis suggests that, under scaling, GR can overcome the inherent limitations of DR, yielding two major benefits. First, with larger corpora, GR avoids the sharp performance degradation caused by the optimization drift induced by DR's local normalization. Second, with larger models, GR's representational capacity scales with parameter size, u

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyzhang2025does

Related papers

• Generative Dense Retrieval: Memory Can Be A Burden (2024)4.52
• Generative Retrieval Meets Multi-graded Relevance (2024)2.26
• Generative Retrieval As Dense Retrieval (2023)0.00
• Continual Learning For Generative Retrieval Over Dynamic Corpora (2023)11.49
• Generative Retrieval As Multi-vector Dense Retrieval (2024)8.60
• Scalable And Effective Generative Information Retrieval (2023)10.48
• Lightweight And Direct Document Relevance Optimization For Generative Information Retrieval (2025)4.52
• Differentiable Geometric Indexing For End-to-end Generative Retrieval (2026)0.00