DiffuGR: Generative Document Retrieval with Diffusion Language Models

TL;DR

DiffuGR introduces a diffusion-based generative document retrieval method that improves accuracy and efficiency control over traditional auto-regressive models, enabling parallel DocID generation and dynamic trade-off management.

Contribution

It proposes a novel diffusion language model approach for document retrieval, addressing limitations of auto-regressive methods with parallel generation and controllable quality-latency trade-offs.

Findings

Outperforms strong auto-regressive retrievers on benchmarks.

Enables explicit control over retrieval quality and latency.

Demonstrates effective parallel DocID generation with diffusion refinement.

Abstract

Generative retrieval (GR) reframes document retrieval as an end-to-end task of generating sequential document identifiers (DocIDs). Existing GR methods predominantly rely on left-to-right auto-regressive decoding, which suffers from two fundamental limitations: (i) a \emph{mismatch between DocID generation and natural language generation}, whereby an incorrect DocID token generated at an early step can lead to entirely erroneous retrieval; and (ii) an \emph{inability to dynamically balance the trade-off between retrieval efficiency and accuracy}, which is crucial for practical applications. To tackle these challenges, we propose generative document retrieval with diffusion language models, termed \emph{DiffuGR}. DiffuGR formulates DocID generation as a discrete diffusion process. During training, DocIDs are corrupted through a stochastic masking process, and a diffusion language model is trained to recover them under a retrieval-aware objective. For inference, DiffuGR generates DocID tokens in parallel and refines them through a controllable number of denoising steps. Unlike auto-regressive decoding, DiffuGR introduce \emph{a novel mechanism to first generate plenty of confident DocID tokens and then refine the generation through diffusion-based denoising}. Moreover, DiffuGR also offers \emph{explicit runtime control over the quality-latency tradeoff}. Extensive experiments on widely-applied retrieval benchmarks show that DiffuGR outperforms strong auto-regressive generative retrievers. Additionally, we verify that DiffuGR achieves flexible control over the quality-latency trade-off via variable denoising budgets.