mlsanityData to 7 Oct 2026

Modality-Guided Mixture of Structured Experts with Entropy-Triggered Routing for Multimodal Recommendation

arXiv 2602.20723 ↗First posted 24 Feb 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Multimodal recommenders combine collaborative behavior with visual and textual item evidence, whose usefulness varies across user-item interactions. Independently trained source-specific diagnostic probes partition held-out interactions into behavior-, appearance-, semantics-, and mixed-evidence regimes across five benchmarks, within which capacity-matched fixed fusion rules exhibit systematic regime-dependent performance crossovers. This diagnostic observation motivates MAGNET, a multimodal graph recommender with two core mechanisms. First, a calibrated expert bank organizes trainable experts by anchor source (behavior, appearance, or semantics) and fusion family (dominant, balanced, or complementary), while an interaction-conditioned router selects a sparse composition using all three evidence sources. Second, an entropy-triggered, coverage-aware progressive schedule decouples population-level routing-mass coverage from per-instance decisiveness, transitioning from broad routing exploration to confident specialization once sufficient coverage is sustained, while preserving coverage thereafter. MAGNET separately encodes the observed interaction graph and a filtered content-induced structural view, applying cross-view alignment after independent propagation. We evaluate MAGNET on four Amazon domains and the non-Amazon MicroLens-100K benchmark, including tail-item and low-history warm-start user evaluation alongside matched expert-design controls. Under the shared-feature, fixed-split protocol, MAGNET-DV exceeds the strongest protocol-compatible non-MAGNET baseline in every reported main-table cell, supported by seed-wise difference tests. Both the fixed and KL-anchored structured variants achieve higher five-seed mean NDCG@20 than matched homogeneous and free-mixture alternatives across all five datasets. MAGNET supports route-level diagnostics through explicit expert semantics.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 31 of its 46 references that are in the map. First posted on arXiv 24 Feb 2026.

Builds on

31 of 46 references.

References · 31

  1. 2017Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layercited by 3,413
  2. 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,014
  3. 2020LightGCN: Simplifying and Powering Graph Convolution Network for Recommendationcited by 1,134
  4. 2024DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Modelscited by 834
  5. 2021Scaling Vision with Sparse Mixture of Expertscited by 740
  6. 2019Neural Graph Collaborative Filteringcited by 660
  7. 2020Self-supervised Graph Learning for Recommendationcited by 305
  8. 2025Reasoning with Exploration: An Entropy Perspectivecited by 277
  9. 2022Balanced Multimodal Learning via On-the-fly Gradient Modulationcited by 259
  10. 2022Multimodal Contrastive Learning with LIMoE: the Language-Image Mixture of Expertscited by 202
  11. 2015VBPR: Visual Bayesian Personalized Ranking from Implicit Feedbackcited by 151
  12. 2020Disentangled Graph Collaborative Filteringcited by 118
19 more
  1. 2024Uni-MoE: Scaling Unified Multimodal LLMs with Mixture of Expertscited by 107
  2. 2022Characterizing and overcoming the greedy nature of learning in multi-modal deep neural networkscited by 92
  3. 2017Joint Deep Modeling of Users and Items Using Reviews for Recommendationcited by 84
  4. 2021Mining Latent Structures for Multimedia Recommendationcited by 79
  5. 2023Graph Mixture of Experts: Learning on Large-Scale Graphs with Explicit Diversity Modelingcited by 63
  6. 2021GRCN: Graph-Refined Convolutional Network for Multimedia Recommendation with Implicit Feedbackcited by 58
  7. 2022Bootstrap Latent Representations for Multi-modal Recommendationcited by 55
  8. 2022A Tale of Two Graphs: Freezing and Denoising Graph Structures for Multimodal Recommendationcited by 53
  9. 2023A Content-Driven Micro-Video Recommendation Dataset at Scalecited by 43
  10. 2024Med-MoE: Mixture of Domain-Specific Experts for Lightweight Medical Vision-Language Modelscited by 43
  11. 2023Multi-View Graph Convolutional Network for Multimedia Recommendationcited by 32
  12. 2024GraphMoRE: Mitigating Topological Heterogeneity via Mixture of Riemannian Expertscited by 23
  13. 2023MMoE: Enhancing Multimodal Models with Mixtures of Multimodal Interaction Expertscited by 21
  14. 2024MENTOR: Multi-level Self-supervised Learning for Multimodal Recommendationcited by 21
  15. 2022Disentangled Multimodal Representation Learning for Recommendationcited by 15
  16. 2024Spectrum-based Modality Representation Fusion Graph Convolutional Network for Multimodal Recommendationcited by 13
  17. 2025COHESION: Composite Graph Convolutional Network with Dual-Stage Fusion for Multimodal Recommendationcited by 9
  18. 2024GUME: Graphs and User Modalities Enhancement for Long-Tail Multimodal Recommendationcited by 8
  19. 2025VI-MMRec: Similarity-Aware Training Cost-free Virtual User-Item Interactions for Multimodal Recommendationcited by 2

Built on by

Not cited yet.