STAIR: Manipulating Collaborative and Multimodal Information for E-Commerce Recommendation

Cong Xu; Jun Wang; Wei Zhang; Yunhang He

REVIEW 3 major objections 4 minor 42 references

STAIR: Manipulating Collaborative and Multimodal Information for E-Commerce Recommendation

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that in e-commerce, collaborative and multimodal signals can coexist in one embedding by giving each dimension its own graph-convolution layer weights, and that this design outperforms previous multimodal recommenders on…

desk verdict STAIR's empirical results are solid and the method is genuinely new, but the paper's central dimension-split mechanism is unverified and needs either evidence or a softer claim. read the letter →

arxiv 2412.11729 v1 pith:7DUX3DNI submitted 2024-12-16 cs.IR

Cong Xu , Yunhang He , Jun Wang , Wei Zhang This is my paper · ORCID

classification cs.IR

keywords multimodalrecommendatione-commercecollaborativefilteringgraphconvolutionalnetworksmodalityerasureforgettingstepwiseconvolutionembeddinginitialization

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that in e-commerce, user behavior is rarely determined by image or text content, so multimodal features should complement rather than replace collaborative signals. It identifies two failure modes: standard graph convolution erases modality information from embeddings, and the recommendation loss gradually forgets it during training. STAIR addresses both by initializing embeddings from whitened multimodal features and applying a stepwise graph convolution that gives different layer weights to different embedding dimensions, letting collaborative and multimodal signals coexist. The authors report that STAIR outperforms prior multimodal and graph-based methods on the Baby, Sports, and Electronics datasets while keeping time and memory costs close to LightGCN.

What carries the argument

The central object is the stepwise graph convolution: a per-dimension, per-layer weight scheme that generalizes LightGCN. The weight of dimension $j$ at layer $l$ is $\alpha_{jl} = (1-\beta_j)/(1-\beta_j^{L+1}) \cdot \beta_j^l$ with $\beta_j = 0.9(1 - ((j-1)/d)^\gamma)$, normalized so the $L+1$ weights sum to one. Early dimensions (with $\beta_j$ close to 1) behave like LightGCN and carry collaborative signal; late dimensions (with $\beta_j$ close to 0) concentrate weight on the initial layer and carry multimodal signal. The backward pass applies the complementary ratios $\beta'_j = 1 - \beta_j$ over a symmetrized kNN graph built from textual and visual features, constraining item-embedding updates so modality-similar items move together; the paper cites a prior convergence result for this constrained update.

What would settle it

Zero out the first half of the embedding dimensions in a trained STAIR model and then the second half, measuring Recall@20 and the Pearson correlation between embeddings and the original multimodal features in each case. If removing late dimensions does not remove most of the modality correlation and removing early dimensions does not remove most of the collaborative performance, the paper's claimed dimension-level separation is not what drives the gains.

Watch

Extended reading notes

Core claim

The central claim is that modality erasure and modality forgetting are the two reasons multimodal features underperform in e-commerce, and both can be fixed inside a single graph-convolution framework. Because LightGCN acts on each embedding dimension independently, STAIR assigns dimension-specific layer weights: dimension $j$ uses teleport ratio $\beta_j = 0.9(1 - ((j-1)/d)^\gamma)$, so early dimensions smooth like LightGCN and absorb collaborative signal while late dimensions stay close to the raw multimodal initialization. On the backward pass, gradients are smoothed with a modality-kNN similarity matrix using the complementary ratio $\beta'_j = 1 - \beta_j$, which makes modality-similar items update together and slows forgetting. Starting from whitened multimodal features, this forward-backward stepwise convolution yields reported state-of-the-art results on all three datasets.

Load-bearing premise

The design depends on the idea that ordering dimensions from collaborative to multimodal has meaning: if the coordinates are interchangeable, the stepwise weighting is just a fixed per-dimension reweighting and the reported gains could come from extra hyperparameters rather than from protecting modality information.

Editorial extensions

If this is right

On all three datasets, STAIR's Recall@10/20 and NDCG@10/20 exceed the strongest prior baseline FREEDOM by roughly 2 to 6 percent relative, and LightGCN by roughly 9 to 24 percent.
STAIR's per-epoch time and GPU memory are close to LightGCN (for example, 1.45 seconds versus 1.25 seconds on Baby, and 490 MB versus 478 MB), while contrastive methods such as MMSSL are far more expensive and cannot run on Electronics under the reported GPU.
Modality initialization alone improves accuracy and reduces sensitivity to random seeds, and the backward stepwise convolution restores the Pearson correlation between trained embeddings and the initial multimodal features over training.
The single hyperparameter $\gamma$ controls the collaborative versus multimodal balance; tuning it lower, which retains more modality information, helps on datasets with higher modal-specific user behavior uncertainty.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the dimension split is real, the late embedding coordinates are a reusable representation of modality-as-complement: one could attach a lightweight modality-aware head to those coordinates without retraining the whole model.
The same dimension-partitioning principle could be applied to other graph filters, not just LightGCN: replacing equal layer weights with a monotone sweep along dimensions may protect other collaborative filters from modality erasure.
The paper's own conclusion suggests a natural stress test: on content-driven datasets such as news or video, the optimal $\gamma$ should shift toward collaborative-heavy settings, and raw multimodal features should matter more; measuring that shift would show whether the method generalizes beyond e-commerce.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

STAIR is a genuinely new method for fusing collaborative and multimodal signals in e-commerce recommendation, and its empirical claims hold up reasonably well. The stepwise graph convolution and the backward constrained update are both new relative to the cited LightGCN, LATTICE, and FREEDOM, and the paper tests them on three Amazon datasets with ablations, hyperparameter sensitivity, and a clear efficiency comparison. The reported gains—2–6% over FREEDOM and 10–20% over LightGCN at roughly LightGCN's cost—are consistent across datasets and metrics, and the code is public. That is a useful contribution for an applied subfield.

The soft spot is that the paper's central explanatory mechanism is not directly verified. FSC assigns each embedding dimension a different teleport ratio, and the text claims early dimensions carry collaborative signal while later dimensions retain multimodal signal. No experiment measures per-dimension content; the figures show aggregate performance and average correlation. If that semantic split is absent, FSC is just a fixed per-dimension weighting, and the gains could come from the extra hyperparameters (γ, k_t, k_v). This is the main thing I would want a revision to address, and it is fixable: a per-dimension analysis or at least a softer claim would do.

Two minor issues. Table 2 reports p-values but no standard deviations; p-values are helpful but std devs let the reader judge effect size. The convergence corollary for BSC is imported from a self-cited NeurIPS paper and looks shaky: since β'_j = 1 − β_j can be very close to 1, the stated O(1/((1 − max β_j)^2 T)) bound does not obviously follow from the cited theorem. That needs a clearer derivation or a corrected statement, but it is not load-bearing for the empirical result.

None of this sinks the paper. The empirical benchmark comparison is honest, the ablations show the components matter, and the efficiency claims are concrete. I would send it to reviewers with a request for the per-dimension evidence or a weakened mechanistic claim, plus the reporting fixes.

For a reader working on multimodal recommendation or efficient graph-based CF, this is worth a careful look. I would cite it if I worked in that area.

Referee Report

3 major / 4 minor

Summary. The paper proposes STAIR, a multimodal recommendation method for e-commerce that combines collaborative and multimodal information while avoiding two identified problems: modality erasure in graph convolution and modality forgetting during training. The method initializes item embeddings from whitened multimodal features, applies a forward stepwise graph convolution (FSC) that assigns per-dimension layer weights so that earlier embedding dimensions emphasize collaborative signal and later dimensions retain multimodal signal, and uses a backward stepwise convolution (BSC) to constrain item embedding updates based on a modality-kNN similarity graph. Experiments on Baby, Sports, and Electronics report state-of-the-art Recall@N and NDCG@N over strong baselines, with relative gains of 2–6% over FREEDOM and large efficiency advantages in time and memory.

Significance. If the central claims hold, STAIR is a practically valuable and unusually efficient multimodal recommender for e-commerce: it achieves consistent gains over contrastive and graph-based multimodal baselines while running at roughly the cost of LightGCN, and it offers a clear conceptual diagnosis (erasure and forgetting) with simple fixes. The paper includes several commendable practices: paired t-tests over five runs, ablation studies, hyperparameter sensitivity analysis, efficiency comparison, and a public code link with detailed hyperparameter tables. The main weakness is that the load-bearing conceptual mechanism—per-dimension separation of collaborative and multimodal information—is asserted rather than directly verified, and one convergence guarantee is imported from the authors' prior work without a proof of the required generalization.

major comments (3)

[Forward Stepwise Convolution (FSC), Eq. (2)] The core design claim that assigning a varying teleport ratio β_j along the embedding dimension makes early dimensions carry collaborative information and later dimensions retain multimodal information is not directly verified anywhere in the paper. Figures 3 and 5 show aggregate performance and average correlation, not per-dimension trends, and Figure 4a only illustrates the assigned weights. If the dimension ordering does not actually induce semantic separation, FSC reduces to a fixed per-dimension weighting of LightGCN-style smoothing, and the reported gains over FREEDOM could be attributable to the additional hyperparameters γ, k_t, k_v rather than to the proposed co-existence mechanism. Please add a direct test of the separation hypothesis, for example by measuring per-dimension correlation with modality initialization after training, by evaluating recommendation quality using only subsets of dimensions, or by comparing against a version with randomly permuted or reversed dimension assignments.
[Overall Performance Comparison, Table 2] The improvements over FREEDOM are modest (2–6%) and the table reports p-values but no standard deviations. Since the best checkpoints are selected on validation NDCG@20 over 500 epochs, run-to-run variance could be non-negligible; reporting means with standard deviations (or confidence intervals) for all metrics would make the SOTA claim substantially more robust. Please add these statistics.
[Backward Stepwise Convolution (BSC), Eq. (4) and Eq. (5)] The convergence statement for BSC is presented as a corollary of Xu et al. (2024), but that prior work proves convergence for a constant teleport ratio β, whereas Eq. (5) uses a per-dimension β'_j = 1 − β_j. The extension to dimension-dependent weights is not obviously trivial, and the update in Eq. (4) also depends on the symmetrized modality similarity matrix S̃. A proof or a precise statement of the conditions under which the corollary holds should be included; as written, the theoretical guarantee is unsupported.

minor comments (4)

[Appendix, User Behavior Uncertainty Details] There is a typo: "simiarity" should be "similarity", and in the main text the spacing in "marked inbold" should be fixed.
[Motivation: Modal-Specific Behavior Uncertainty] The main text refers to Figure 2 and later to Figure 8 for the same uncertainty results; the figure numbering should be made consistent and the reference should point to the appropriate figure at first mention.
[Related Work and Introduction] The claim of being "the first to identify" poor modal-specific behavior uncertainty in e-commerce is strong given that prior works (e.g., LATTICE and FREEDOM) already discuss the limited utility of raw modalities in this scenario; consider softening the novelty claim to "quantitatively justify".
[Training and Prediction, Eq. (6)] The notation i′̸∈Nu in Eq. (6) is rendered with a broken prime symbol and should be typeset cleanly as i′ ∉ N_u.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: STAIR's SOTA claim rests on external benchmarks; FSC and BSC are explicit design mechanisms, and the only self-citation (BSC convergence from Xu et al. 2024) is not load-bearing for the central empirical result.

full rationale

STAIR's central claim is an empirical performance comparison on Baby, Sports, and Electronics against external baselines such as FREEDOM, LATTICE, and LightGCN, so the headline result is not fitted into existence by the paper's own definitions. The forward stepwise convolution is a parameterized per-dimension interpolation between LightGCN-style propagation and the identity (Eqs. (1)-(2)); its stated effect, that earlier dimensions are more smoothed while later dimensions stay closer to the multimodal initialization, is a direct property of the chosen weights rather than a circularly derived prediction. The backward stepwise convolution (Eq. (4)) is similarly an explicit update rule built from a modality-similarity graph; the paper validates it with external recommendation metrics and an ablation, not by assuming the conclusion. The only self-citation is the convergence guarantee borrowed from Xu et al. 2024, which is a supporting theoretical remark; the experiments themselves demonstrate stable training, so this citation is not load-bearing for the paper's empirical contribution. The unverified dimension-wise semantic separation in FSC is a correctness or interpretability risk, not a circularity, because the paper does not claim to derive that separation from an independent premise; it builds the separation into the weight schedule. No step in the derivation reduces to its own input by construction.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim depends on a few tuned hyperparameters and on two domain assumptions about multimodal features. The main load-bearing assumptions are the meaningfulness of the dimension ordering in FSC and the transfer of the convergence theorem from constant beta to per-dimension beta. No new physical or conceptual entities are introduced.

free parameters (5)

gamma (stepwise exponent) = Baby 0.1, Sports 0.2, Electronics 0.4
Controls how quickly layer weights shift from collaborative to multimodal across embedding dimensions; selected by validation NDCG@20 from {0.01, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5}.
k_t (text kNN neighbors) = 5 (all datasets)
Number of text-modality neighbors used to build the item similarity graph; selected from {1, 3, 5, 10, 20}.
k_v (visual kNN neighbors) = 1 (all datasets)
Number of visual-modality neighbors used to build the item similarity graph; selected from {1, 3, 5, 10, 20}.
learning rate = 1e-3 (all datasets)
Selected from {1e-4, 5e-4, 1e-3, 5e-3}; standard optimization parameter that affects the final results.
weight decay = Baby 0.3, Sports 0.1, Electronics 0.1
Selected in [0, 1]; standard regularization parameter that affects the final results.

assumptions (4)

domain assumption Raw multimodal features, after whitening, provide a useful embedding initialization.
Section 'Motivation' says the whitening technique is used because there is no well-trained projector; this assumes the top-d left singular vectors preserve enough multimodal signal.
domain assumption Cosine-similarity kNN graphs on raw modalities capture relevant item-item similarity.
Used in Eq. (3) to define neighbors and similarity matrix S; follows the approach of LATTICE and FREEDOM.
ad hoc to paper Per-dimension layer weights in FSC can separate collaborative and multimodal information by embedding coordinate.
Central design premise of FSC; no direct per-dimension proof is provided.
ad hoc to paper The convergence guarantee for BSC with constant beta extends to per-dimension beta'_j.
The paper cites Xu et al. 2024 for constant beta and states a corollary giving O(1/((1 - max_j beta_j)^2 T)); the extension is not derived here.

how reviews work

0 comments

Cite this review

Pith. "Pith review of STAIR: Manipulating Collaborative and Multimodal Information for E-Commerce Recommendation." pith.science (2026). https://pith.science/paper/7DUX3DNI

@misc{pith2026241211729,
  author       = {Pith},
  title        = {Pith review of: STAIR: Manipulating Collaborative and Multimodal Information for E-Commerce Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7DUX3DNI}},
  note         = {Machine review of arXiv:2412.11729}
}

read the original abstract

While the mining of modalities is the focus of most multimodal recommendation methods, we believe that how to fully utilize both collaborative and multimodal information is pivotal in e-commerce scenarios where, as clarified in this work, the user behaviors are rarely determined entirely by multimodal features. In order to combine the two distinct types of information, some additional challenges are encountered: 1) Modality erasure: Vanilla graph convolution, which proves rather useful in collaborative filtering, however erases multimodal information; 2) Modality forgetting: Multimodal information tends to be gradually forgotten as the recommendation loss essentially facilitates the learning of collaborative information. To this end, we propose a novel approach named STAIR, which employs a novel STepwise grAph convolution to enable a co-existence of collaborative and multimodal Information in e-commerce Recommendation. Besides, it starts with the raw multimodal features as an initialization, and the forgetting problem can be significantly alleviated through constrained embedding updates. As a result, STAIR achieves state-of-the-art recommendation performance on three public e-commerce datasets with minimal computational and memory costs. Our code is available at https://github.com/yhhe2004/STAIR.

Figures

Figures reproduced from arXiv: 2412.11729 by the authors.

**Figure 1.** Overview of STAIR. 1) Forward pass: More multimodal information are retained in the later dimensions via FSC. 2) Backward pass: BSC is used to enhance multimodal information to mitigate modality forgetting problem. 0.0 0.5 1.0 1.5 Uncertainty MF-BPR Textual Visual Random 0.7846 1.1168 1.1783 1.2633 Baby 0.0 0.5 1.0 1.5 Uncertainty 0.5626 0.8346 1.1631 1.2577 Sports [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** User behavior uncertainty under various features. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Modality erasure problem. Multimodal infor [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 5.** Figure 5: Modality forgetting problem. Pearson correlation [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

**Figure 6.** Figure 6: STAIR ablation experiments. 0.01 0.1 0.2 0.3 0.5 γ 0.090 0.095 0.100 0.105 Recall@20 NDCG@20 0.040 0.045 0.050 1 3 5 10 20 kv 20 10 5 3 1 kt 0.043 0.044 0.044 0.044 0.043 0.045 0.046 0.045 0.043 0.041 0.045 0.045 0.044 0.041 0.040 0.044 0.043 0.042 0.041 0.040 0.042 0.…

**Figure 7.** Figure 7: shows the hyperparameter sensitivity on the Baby dataset. 1) According to Eq. (2), a larger γ implies more em- 0.08 0.10 Recall@20 STAIR w/o MI w/o FSC w/o BSC LightGCN Baby Sports 0.03 0.04 0.05 NDCG@20 [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

**Figure 8.** Figure 8: User behavior uncertainty under various features. [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 34 canonical work pages

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

Arthur, D.; and Vassilvitskii, S. 2006. k-means++: The advantages of careful seeding. Technical report, Stanford

2006
[4]

Berg, R. v. d.; Kipf, T. N.; and Welling, M. 2017. Graph convolutional matrix completion. arXiv preprint arXiv:1706.02263

arXiv 2017
[5]

Chen, C.; Zhang, M.; Zhang, Y.; Liu, Y.; and Ma, S. 2020. Efficient neural matrix factorization without sampling for recommendation. ACM Transactions on Information Systems (TOIS), 38(2): 1--28

work page 2020
[6]

Chen, Q.; Zhao, H.; Li, W.; Huang, P.; and Ou, W. 2019. Behavior sequence transformer for e-commerce recommendation in alibaba. In International Workshop on Deep Learning Practice for High-dimensional Sparse Data, 1--4

work page 2019
[7]

El-Kishky, A.; Markovich, T.; Park, S.; Verma, C.; Kim, B.; Eskander, R.; Malkov, Y.; Portman, F.; Samaniego, S.; Xiao, Y.; et al. 2022. Twhin: Embedding the twitter heterogeneous information network for personalized recommendation. In International Conference on Knowledge Discovery & Data Mining (KDD), 2842--2850

work page 2022
[8]

Gasteiger, J.; Bojchevski, A.; and G \"u nnemann, S. 2018. Predict then propagate: Graph neural networks meet personalized pagerank. In International Conference on Learning Representations (ICLR)

work page 2018

Show all 42 references

[9]

Guo, J.; Du, L.; Chen, X.; Ma, X.; Fu, Q.; Han, S.; Zhang, D.; and Zhang, Y. 2023. On Manipulating Signals of User-Item Graph: A Jacobi Polynomial-based Graph Collaborative Filtering. In International Conference on Knowledge Discovery & Data Mining (KDD), 602--613

2023
[10]

He, R.; and McAuley, J. 2015. VBPR: Visual bayesian personalized ranking from implicit feedback

2015
[11]

He, R.; and McAuley, J. 2016. Ups and downs: Modeling the visual evolution of fashion trends with one-class collaborative filtering. In International Conference on World Wide Web (WWW), 507--517

2016
[12]

He, X.; Deng, K.; Wang, X.; Li, Y.; Zhang, Y.; and Wang, M. 2020. Lightgcn: Simplifying and powering graph convolution network for recommendation. In International Conference on Research and Development in Information Retrieval (SIGIR), 639--648

2020
[13]

He, X.; Liao, L.; Zhang, H.; Nie, L.; Hu, X.; and Chua, T.-S. 2017. Neural collaborative filtering. In International Conference on World Wide Web (WWW), 173--182

2017
[14]

Hoffmann, H. 2007. Kernel PCA for novelty detection. Pattern Recognition (PR), 40(3): 863--874

2007
[15]

Hu, H.; Liu, Q.; Li, C.; and Kan, M.-Y. 2024. Lightweight modality adaptation to sequential recommendation via correlation supervision. In European Conference on Information Retrieval (ECIR), 123--139

2024
[16]

Huang, J.; Tang, D.; Zhong, W.; Lu, S.; Shou, L.; Gong, M.; Jiang, D.; and Duan, N. 2021 a . WhiteningBERT: An easy unsupervised sentence embedding approach. In Findings of Conference on Empirical Methods in Natural Language Processing (EMNLP), 238--244

2021
[17]

Huang, Q.; He, H.; Singh, A.; Lim, S.-N.; and Benson, A. 2021 b . Combining label propagation and simple models out-performs graph neural networks. In International Conference on Learning Representations (ICLR)

2021
[18]

N.; and Welling, M

Kipf, T. N.; and Welling, M. 2016. Semi-supervised classification with graph convolutional networks. In International Conference on Learning Representations (ICLR)

2016
[19]

Koren, Y. 2008. Factorization meets the neighborhood: a multifaceted collaborative filtering model. In International Conference on Knowledge Discovery & Data Mining (KDD), 426–434

2008
[20]

Li, Q.; Han, Z.; and Wu, X.-M. 2018. Deeper insights into graph convolutional networks for semi-supervised learning. In Conference on Artificial Intelligence (AAAI)

2018
[21]

McElfresh, D.; Khandagale, S.; Valverde, J.; Dickerson, J.; and White, C. 2022. On the generalizability and predictability of recommender systems. In Advances in Neural Information Processing Systems (NeurIPS), volume 35, 4416--4432

2022
[22]

Ni, J.; Li, J.; and McAuley, J. 2019. Justifying recommendations using distantly-labeled reviews and fine-grained aspects. In Conference on Empirical Methods in Natural Language Processing and International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 188--197

2019
[23]

Reimers, N.; and Gurevych, I. 2019. Sentence-BERT: Sentence embeddings using siamese bert-networks. In Conference on Empirical Methods in Natural Language Processing (EMNLP)

2019
[24]

Rendle, S.; Freudenthaler, C.; Gantner, Z.; and Schmidt-Thieme, L. 2009. BPR: Bayesian personalized ranking from implicit feedback. In Uncertainty in Artificial Intelligence (UAI)

2009
[25]

K.; and Li, D

Shen, Y.; Wu, Y.; Zhang, Y.; Shan, C.; Zhang, J.; Letaief, B. K.; and Li, D. 2021. How powerful is graph convolution for recommendation? In International Conference on Information & Knowledge Management (CIKM), 1619--1629

2021
[26]

Su, J.; Cao, J.; Liu, W.; and Ou, Y. 2021. Whitening sentence representations for better semantics and faster retrieval. arXiv preprint arXiv:2103.15316

2021 arXiv
[27]

Wang, F.; Liu, Y.; Liu, K.; Wang, Y.; Medya, S.; and Yu, P. S. 2024. Uncertainty in graph neural networks: A survey. arXiv preprint arXiv:2403.07185

2024 arXiv
[28]

Wang, Q.; Wei, Y.; Yin, J.; Wu, J.; Song, X.; and Nie, L. 2021. Dualgnn: Dual graph neural network for multimedia recommendation. IEEE Transactions on Multimedia, 25: 1074--1084

2021
[29]

Wang, X.; He, X.; Wang, M.; Feng, F.; and Chua, T.-S. 2019. Neural Graph Collaborative Filtering. In International Conference on Research and Development in Information Retrieval (SIGIR), SIGIR ’19

2019
[30]

Wei, W.; Huang, C.; Xia, L.; and Zhang, C. 2023 a . Multi-modal self-supervised learning for recommendation. In Proceedings of the ACM Web Conference (WWW), 790--800

2023
[31]

Wei, Y.; Liu, W.; Liu, F.; Wang, X.; Nie, L.; and Chua, T.-S. 2023 b . Lightgt: A light graph transformer for multimedia recommendation. In International Conference on Research and Development in Information Retrieval (SIGIR), 1508--1517

2023
[32]

Wei, Y.; Wang, X.; Nie, L.; He, X.; Hong, R.; and Chua, T.-S. 2019. MMGCN: Multi-modal graph convolution network for personalized recommendation of micro-video. In International Conference on Multimedia (MM), 1437--1445

2019
[33]

Wu, F.; Qiao, Y.; Chen, J.-H.; Wu, C.; Qi, T.; Lian, J.; Liu, D.; Xie, X.; Gao, J.; Wu, W.; et al. 2020. Mind: A large-scale dataset for news recommendation. In Annual Meeting of the Association for Computational Linguistics (ACL), 3597--3606

2020
[34]

Wu, F.; Souza, A.; Zhang, T.; Fifty, C.; Yu, T.; and Weinberger, K. 2019. Simplifying graph convolutional networks. In International Conference on Machine Learning (ICML), 6861--6871

2019
[35]

Xu, C.; Wang, J.; Wang, J.; and Zhang, W. 2024. Graph-enhanced optimizers for structure-aware recommendation embedding evolution. In Advances in Neural Information Processing Systems (NeurIPS)

2024
[36]

Yang, M.; Li, Z.; Zhou, M.; Liu, J.; and King, I. 2022. Hicf: Hyperbolic informative collaborative filtering. In Conference on Knowledge Discovery & Data Mining (KDD), 2212--2221

2022
[37]

Yu, P.; Tan, Z.; Lu, G.; and Bao, B.-K. 2023. Multi-view graph convolutional network for multimedia recommendation. In International Conference on Multimedia (MM), 6576--6585

2023
[38]

Zhang, J.; Zhu, Y.; Liu, Q.; Wu, S.; Wang, S.; and Wang, L. 2021. Mining latent structures for multimedia recommendation. In International Conference on Multimedia (MM), 3872--3880

2021
[39]

Zhou, X. 2023. MMRec: Simplifying multimodal recommendation. arXiv preprint arXiv:2302.03497

2023 arXiv
[40]

Zhou, X.; and Shen, Z. 2023. A tale of two graphs: Freezing and denoising graph structures for multimodal recommendation. In International Conference on Multimedia (MM), 935--943

2023
[41]

Zhou, X.; Zhou, H.; Liu, Y.; Zeng, Z.; Miao, C.; Wang, P.; You, Y.; and Jiang, F. 2023. Bootstrap latent representations for multi-modal recommendation. In Proceedings of the ACM Web Conference (WWW). ACM

2023
[42]

Zhu, X.; Jin, M.; Zhang, H.; Meng, C.; Zhang, D.; and Li, X. 2024. Modeling Domains as Distributions with Uncertainty for Cross-Domain Recommendation. In International Conference on Research and Development in Information Retrieval (SIGIR), 2517--2521

2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

Arthur, D.; and Vassilvitskii, S. 2006. k-means++: The advantages of careful seeding. Technical report, Stanford

2006

[4] [4]

Berg, R. v. d.; Kipf, T. N.; and Welling, M. 2017. Graph convolutional matrix completion. arXiv preprint arXiv:1706.02263

arXiv 2017

[5] [5]

Chen, C.; Zhang, M.; Zhang, Y.; Liu, Y.; and Ma, S. 2020. Efficient neural matrix factorization without sampling for recommendation. ACM Transactions on Information Systems (TOIS), 38(2): 1--28

work page 2020

[6] [6]

Chen, Q.; Zhao, H.; Li, W.; Huang, P.; and Ou, W. 2019. Behavior sequence transformer for e-commerce recommendation in alibaba. In International Workshop on Deep Learning Practice for High-dimensional Sparse Data, 1--4

work page 2019

[7] [7]

El-Kishky, A.; Markovich, T.; Park, S.; Verma, C.; Kim, B.; Eskander, R.; Malkov, Y.; Portman, F.; Samaniego, S.; Xiao, Y.; et al. 2022. Twhin: Embedding the twitter heterogeneous information network for personalized recommendation. In International Conference on Knowledge Discovery & Data Mining (KDD), 2842--2850

work page 2022

[8] [8]

Gasteiger, J.; Bojchevski, A.; and G \"u nnemann, S. 2018. Predict then propagate: Graph neural networks meet personalized pagerank. In International Conference on Learning Representations (ICLR)

work page 2018