Diagnosing Modality Reliability in Sparse E-Commerce Recommendation: A White-Box Routing Study

Multimodal recommenders usually assume that fusing collaborative, visual, and textual signals improves accuracy. We treat this as a diagnostic question rather than an accuracy contest. Under interaction-level leave-one-out evaluation on three sparse Amazon categories, three findings emerge. First, text is the only modality that transfers; removing it lowers AUC by 0.23–0.25, while removing the visual stream changes AUC by at most 0.01 and removing collaborative ID features leaves it unchanged. Second, the apparent value of collaborative features is protocol-dependent. Full models reach AUC 0.97–0.99 under random splitting but only 0.71–0.74 under leave-one-out. The gap vanishes under a user-disjoint split, which is consistent with identity-side shortcuts rather than transferable signal. Third, white-box routing agrees with modality ablations at the dataset level, but is not seed-stable enough to serve as a per-instance explanation. We introduce counterfactual modality masking, which yields signed per-item modality attributions and shows that a modality can help on average while acting as noise for a specific item; gradient-based attribution, which aggregates absolute magnitudes, cannot express this sign. Against recent 2025–2026 multimodal models re-implemented under the identical protocol, LCFC is significantly more accurate than OrthoRec and statistically indistinguishable from I3MRec. Its contribution is not a leaderboard claim but a leakage-aware evaluation protocol together with an auditable diagnostic instrument. All deltas carry five-seed dispersion and paired significance tests.

Authors

Institutions

Publication Details

Journal
Electronics
Published
2026-11-16
DOI
https://doi.org/10.3390/electronics15225334
Primary Topic
Recommender Systems and Techniques
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Diagnosing Modality Reliability in Sparse E-Commerce Recommendation: A White-Box Routing Study

Weiqi Zhang, Guisheng Yin, Wengui Han, Jihui Zhang et al.
1 links
Electronics
Recommender Systems and Techniques
article

Diagnosing Modality Reliability in Sparse E-Commerce Recommendation: A White-Box Routing Study

Weiqi Zhang, Guisheng Yin, Wengui Han, Jihui Zhang, Fukun Chen, Zheng Chen
article en

Abstract

Multimodal recommenders usually assume that fusing collaborative, visual, and textual signals improves accuracy. We treat this as a diagnostic question rather than an accuracy contest. Under interaction-level leave-one-out evaluation on three sparse Amazon categories, three findings emerge. First, text is the only modality that transfers; removing it lowers AUC by 0.23–0.25, while removing the visual stream changes AUC by at most 0.01 and removing collaborative ID features leaves it unchanged. Second, the apparent value of collaborative features is protocol-dependent. Full models reach AUC 0.97–0.99 under random splitting but only 0.71–0.74 under leave-one-out. The gap vanishes under a user-disjoint split, which is consistent with identity-side shortcuts rather than transferable signal. Third, white-box routing agrees with modality ablations at the dataset level, but is not seed-stable enough to serve as a per-instance explanation. We introduce counterfactual modality masking, which yields signed per-item modality attributions and shows that a modality can help on average while acting as noise for a specific item; gradient-based attribution, which aggregates absolute magnitudes, cannot express this sign. Against recent 2025–2026 multimodal models re-implemented under the identical protocol, LCFC is significantly more accurate than OrthoRec and statistically indistinguishable from I3MRec. Its contribution is not a leaderboard claim but a leakage-aware evaluation protocol together with an auditable diagnostic instrument. All deltas carry five-seed dispersion and paired significance tests.

ElectronicsVol. 15(22)
Harbin Engineering University (CN), Liaoning Normal University (CN), Changzhou Institute of Technology (CN)
Department of Education of Liaoning Province, Changzhou Institute of Technology
Recommender Systems and Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon