Unlocking the Potentials of Retrieval-Augmented Generation for Diffusion Language Models

TL;DR

This paper explores integrating Retrieval-Augmented Generation with Diffusion Language Models, identifies semantic drift issues, and proposes a novel method to improve semantic fidelity and generation precision.

Contribution

It systematically evaluates RAG with DLMs, identifies the semantic drift problem, and introduces SPREAD, a query-relevance-guided denoising strategy to enhance semantic alignment.

Findings

RAG with DLMs shows promising potential but has limited precision.

Semantic Drift (RSD) causes deviation from query semantics during generation.

SPREAD effectively reduces RSD and improves answer precision.

Abstract

Diffusion Language Models (DLMs) have recently demonstrated remarkable capabilities in natural language processing tasks. However, the potential of Retrieval-Augmented Generation (RAG), which shows great successes for enhancing large language models (LLMs), has not been well explored, due to the fundamental difference between LLM and DLM decoding. To fill this critical gap, we systematically test the performance of DLMs within the RAG framework. Our findings reveal that DLMs coupled with RAG show promising potentials with stronger dependency on contextual information, but suffer from limited generation precision. We identify a key underlying issue: Response Semantic Drift (RSD), where the generated answer progressively deviates from the query's original semantics, leading to low precision content. We trace this problem to the denoising strategies in DLMs, which fail to maintain semantic alignment with the query throughout the iterative denoising process. To address this, we propose Semantic-Preserving REtrieval-Augmented Diffusion (SPREAD), a novel framework that introduces a query-relevance-guided denoising strategy. By actively guiding the denoising trajectory, SPREAD ensures the generation remains anchored to the query's semantics and effectively suppresses drift. Experimental results demonstrate that SPREAD significantly enhances the precision and effectively mitigates RSD of generated answers within the RAG framework.