Adaptive Diffusion-based Augmentation for Recommendation

TL;DR

This paper introduces ADAR, a diffusion-based module that synthesizes controllable negative samples for recommendation systems, improving their training by generating challenging negatives that enhance model performance.

Contribution

We propose ADAR, a novel diffusion-based method that generates controllable negative samples for recommendation systems, addressing mislabeling issues and improving training effectiveness.

Findings

ADAR significantly improves recommendation accuracy across multiple models.

The method effectively controls negative sample hardness through diffusion transition points.

Experimental results show broad compatibility and substantial performance boosts.

Abstract

Recommendation systems often rely on implicit feedback, where only positive user-item interactions can be observed. Negative sampling is therefore crucial to provide proper negative training signals. However, existing methods tend to mislabel potentially positive but unobserved items as negatives and lack precise control over negative sample selection. We aim to address these by generating controllable negative samples, rather than sampling from the existing item pool. In this context, we propose Adaptive Diffusion-based Augmentation for Recommendation (ADAR), a novel and model-agnostic module that leverages diffusion to synthesize informative negatives. Inspired by the progressive corruption process in diffusion, ADAR simulates a continuous transition from positive to negative, allowing for fine-grained control over sample hardness. To mine suitable negative samples, we theoretically identify the transition point at which a positive sample turns negative and derive a score-aware function to adaptively determine the optimal sampling timestep. By identifying this transition point, ADAR generates challenging negative samples that effectively refine the model's decision boundary. Experiments confirm that ADAR is broadly compatible and boosts the performance of existing recommendation models substantially, including collaborative filtering and sequential recommendation, without architectural modifications.