TimeRoute: Time-Aware Modality Routing and Diffusion for Multi-Modal Recommendation

2026-08-11

AI Summary

This work addresses the challenge in multimodal recommendation that the effectiveness of individual modalities dynamically evolves over time, rendering conventional fixed or globally shared fusion weights suboptimal and prone to introducing outdated noise. To this end, the authors propose a unified diffusion-based recommendation framework that incorporates a time-aware, personalized modality routing mechanism to dynamically generate user-specific modality distributions. Furthermore, they design a FiLM-based dual-stream denoising diffusion graph reconstructor to effectively suppress the influence of obsolete modality edges. This approach is the first to integrate time-aware modality routing with diffusion graph reconstruction, eschewing global fusion strategies to enable precise modeling and filtering of dynamic multimodal signals. Extensive experiments on TikTok, Amazon-Baby, and Amazon-Sports datasets demonstrate consistent improvements over strong baselines, with average gains of up to 9.8% in Recall@K, Precision@K, and NDCG@K across ten random seeds.

Abstract

Multi-modal recommenders fuse collaborative signals with item modalities such as text, images, and audio, but the usefulness of each drifts over time and at different rates. For example, chocolate purchases typically guided by textual ingredient cues can shift toward visual packaging and ambient audio around Valentine's Day. This modality time-scale mismatch gives rise to two coupled challenges: (1) users require different modality proportions across temporal contexts, and (2) less relevant modalities are more likely to introduce outdated or misleading signals into the recommender. We address both challenges within a unified diffusion-based recommender, TimeRoute. A temporal-aware modal router maps each user's aggregated behavioral features to a personalized modality distribution, replacing the globally shared fusion weights used in prior work. The diffusion-based graph reconstructor is then conditioned on the same temporal profile through Feature-wise Linear Modulation (FiLM) with dual-stream long- and short-term denoising heads, suppressing outdated modality edges before they enter the propagation graph. Experiments on TikTok, Amazon-Baby, and Amazon-Sports demonstrate consistent improvements of up to 9.8\% in Recall@K, Precision@K, and NDCG@K over strong baselines across 10-seed paired tests. Code is available at https://anonymous.4open.science/r/TimeRoute.

Problem

• multi-modal recommendation
• temporal dynamics
• modality relevance
• time-aware routing
• outdated signals

Innovation

• time-aware modality routing
• diffusion-based recommendation
• Feature-wise Linear Modulation
• temporal context
• multi-modal fusion