Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

ContextGNN: Beyond Two-Tower Recommendation Systems

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A fused GNN architecture that scores familiar items with pair-wise representations and distant items with a shallow two-tower fallback outperforms both recommendation paradigms on every tested task.

desk verdict A sensible hybrid GNN recommender with a confounded headline result: the 20% gain over NBFNet mostly reflects added candidate coverage, not yet the learned fallback. read the letter →

arxiv 2411.19513 v1 pith:6NDFDNML submitted 2024-11-29 cs.IR cs.LG

Yiwen Yuan , Zecheng Zhang , Xinwei He , Akihiro Nitta , Weihua Hu , Dong Wang , Manan Shah , Shenyang Huang
show 5 more authors
Blaž Stojanovič Alan Krumholz Jan Eric Lenssen Jure Leskovec Matthias Fey
This is my paper · ORCID
classification cs.IRcs.LG
keywords recommendationsystemsgraphneuralnetworkslinkpredictiontwo-towermodelspair-wiserepresentationstemporalheterogeneousgraphslocalityscore
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

Recommendation systems typically use two-tower architectures, whose user and item embeddings are pair-agnostic and miss fine-grained context such as repeat purchases, or pair-wise models, which score only items in a user's local neighborhood and miss exploratory items. This paper argues that the two paradigms are complementary and that a single-stage architecture can fuse them. ContextGNN computes pair-wise scores for items in a user's $k$-hop subgraph, scores everything outside that subgraph with the inner product of a user GNN embedding and a shallow item embedding, and learns a per-user offset that decides how much weight familiar items should receive. On the relational recommendation benchmark, the fused model is reported to improve Mean Average Precision by 20% over the best pair-wise baseline and by 344% over the best two-tower baseline, while needing only one GNN forward pass per user. The significance is that a recommender can be both expressive about local context and global in candidate coverage without a separate candidate-generation stage.

What carries the argument

The load-bearing object is the score-fusion equation that uses pair-wise scores for in-subgraph items and a two-tower score for everything else. The pair-wise branch extends the path-based pair-wise baseline: a $k$-hop subgraph is sampled around the user, an indicator vector is added so the GNN knows which node is the seed user, and both user and item representations are read out from the same forward pass. The two-tower branch is deliberately shallow on the item side: instead of a second item GNN, it uses a plain embedding matrix $W$, which is also injected into the user's GNN input so the user embedding aligns with item embeddings. A per-user MLP on the user embedding produces an additive offset that effectively learns how much weight a user puts on familiar versus exploratory items. Because both branches share the same user subgraph and GNN, the whole model needs only one GNN forward pass per user, giving $\mathcal{O}(|L|)$ complexity for bounded subgraph sizes. The paper also introduces a locality score, the fraction of future ground-truth items inside the $k$-hop neighborhood, to explain when the two-tower fallback matters most.

What would settle it

If, on a cold-start-heavy task with low locality, replacing the shallow item embeddings with a deep item-side GNN produces a higher MAP while everything else is kept equal, the fallback assumption fails. A concrete experiment would hold out a large fraction of items until test time, rank them with both models, and compare MAP restricted to those never-seen items.

Watch

Extended reading notes

Core claim

The central claim is that the two dominant recommendation paradigms answer different questions, and the right model is a weighted sum of both. For each user $v$, ContextGNN samples the $k$-hop neighborhood and runs a bidirectional GNN whose seed is marked with an identifier vector; reading out both the user and item representations yields a pair-wise score $y^{(pair)}_{(v,w)}$ for every item $w$ inside the subgraph. For items outside, the same user representation is matched against a shallow learned embedding $w_w$ to produce a two-tower score $y^{(tower)}_{(v,w)} = h^{(k)}_v \cdot w_w$. The final score is $y_{(v,w)} = y^{(pair)}_{(v,w)} + \mathrm{MLP}_\theta(h^{(k)}_v)$ when $w$ is local and $y^{(tower)}_{(v,w)}$ otherwise. The paper's finding is that this decomposition contextualizes the familiar while falling back to shallow embeddings for the distant, adapting automatically to data with very different locality scores, and that the learned per-user offset captures how exploratory a user is. On every reported relational benchmark task, ContextGNN is the top method, with average MAP gains of 20% over the best pair-wise baseline and 344% over the best two-tower baseline.

Load-bearing premise

The argument rests on the fallback: for an item outside the user's local subgraph, the inner product of a user GNN embedding and a plain shallow item embedding is a strong enough ranking signal, so no item-side GNN or pair-wise context is needed for exploratory items.

Editorial extensions

If this is right

  • ContextGNN ranks first on every task in the relational recommendation benchmark, with average MAP gains of 20% over the best pair-wise baseline and 344% over the best two-tower baseline.
  • Because both score families come out of the same user subgraph and GNN, the model is faster than two-tower GNNs and can train against roughly one million negatives on commodity GPU memory.
  • The gap over the pair-wise baseline grows as the locality score falls, confirming that the shallow two-tower fallback carries the ranking load exactly where local context offers little coverage.
  • The learned per-user fusion offset adapts the model to repeat-oriented versus exploratory users, which the paper shows matters across datasets with different locality characteristics.
  • On a sequential next-item task, the same architecture improves HitRate@1 from 0.148 for the best baseline to 0.411, suggesting the hybrid transfers beyond relational setups.

Reading between the lines

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

  • A natural stress test is cold-start items: swap the shallow item embedding matrix for an item feature encoder and measure whether the gains survive when many target items were never seen in training; the paper names this extension but does not run it.
  • The empirical link between locality score and improvement suggests a deployment-time recipe for choosing subgraph depth $k$ and fusion strength per dataset, but the paper does not provide such a tuning procedure.
  • If the fusion offset truly encodes repeat-versus-explore tendency, it could be extracted as a lightweight user-behavior signal for downstream tasks such as engagement or churn modeling, which the paper does not explore.
  • The near-doubling on the site-sponsor task, where both components already score well, hints that the two score families encode complementary rather than redundant signals; isolating when this complementarity appears would sharpen the design rules for hybrid recommenders.
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.

Referee Report

4 major / 6 minor

Summary. The paper proposes ContextGNN, a hybrid GNN architecture for temporal recommendation that combines (i) a pair-wise, NBFNet-style local scoring of items in a user's k-hop subgraph with (ii) a two-tower fallback that ranks all remaining items via inner products of a GNN user representation with shallow item embeddings, and (iii) a learned per-user offset produced by an MLP that fuses the two score types. The method is evaluated on RelBench recommendation tasks, Amazon-Book static link prediction, and IJCAI next-item prediction, reporting a 20% average MAP improvement over NBFNet and 344% over the best two-tower baseline, while requiring a single GNN forward pass per user.

Significance. If the reported results hold, ContextGNN provides a practical answer to a well-known limitation of pair-wise local models—their inability to recommend candidates outside the local subgraph—without the complexity of multi-stage candidate generation. The locality-score metric (Eq. 1) is a useful diagnostic for characterizing recommendation datasets, and the design observation that item-side GNNs add little on dense item graphs is supported by the SHALLOW ITEM ablation. The authors ship code and evaluate on a standard benchmark (RelBench), and the comparison is strengthened by sharing the same GNN backbone across all GNN baselines. However, the evidence as presented does not yet isolate the mechanism behind the reported gains, and some claims in the abstract go beyond what the tables show.

major comments (4)
  1. [Sec. 4.3, Eq. (2)] As printed, the condition "w ∈ ˜V ∪ R" in Eq. (2) is satisfied by every item w, since R is the full item set; the two-tower fallback branch is therefore never reached and the formal definition contradicts the surrounding text and Sec. 4.2. The intended condition is presumably "w ∈ ˜V ∩ R" for the pair-wise score and "w ∈ R \ ˜V" for the two-tower score. Please correct this equation and ensure the notation for "inside" vs "outside" the subgraph is consistent throughout.
  2. [Sec. 5.2, Table 3; Abstract] The abstract states that ContextGNN "outperforms existing methods, both traditional and GNN-based," but on Amazon-Book (Table 3) UltraGCN reaches Recall@20 0.0681 vs 0.0451 and LightGCL 0.0585 vs 0.0451; ContextGNN is also behind on NDCG@20. The conclusion's "consistently improved upon the state-of-the-art" is similarly too broad. Please either restrict the headline claims to the RelBench temporal setting where ContextGNN is best, or add qualifying language that acknowledges the static small-graph setting as a limitation.
  3. [Sec. 5.1, Table 2; Sec. 4.3] The claimed 20% average gain over NBFNet is confounded with candidate-set coverage: NBFNet is restricted to ranking items in the k-hop subgraph, while ContextGNN additionally scores all items outside via the two-tower fallback. The ablation compares ContextGNN against NBFNet and against SHALLOW ITEM alone, but there is no baseline that supplements NBFNet with a simple full-corpus fallback (e.g., item popularity, or NBFNet local scores plus shallow-item scores without the learned fusion offset). Without such a baseline, the specific contribution of the learned fallback and personalized fusion score is not isolated. To support the claim that the hybrid fusion "is able to adapt," please add an ablation with a non-learned or minimal fallback over the full item corpus.
  4. [Sec. 5, Tables 2–4] No error bars, multiple seeds, or significance tests are reported for any of the main results. Several differences are small in absolute terms (e.g., 2.71 vs 2.60 on rel-amazon val; 15.22 vs 15.17 on rel-stack val) and may be within run-to-run noise. Please report mean and standard deviation over at least three seeds, and provide a significance statement for the headline 20% and 344% average improvements.
minor comments (6)
  1. [Sec. 3, Eq. (1)] Equation (1) contains a stray "=" after the summation symbol; the definition should read s = (1/|L|) Σ_v |N ∩ R ∩ Y| / |Y| without the extra equals sign.
  2. [Sec. 1] In the sentence describing quadratic complexity, "ineffable" should be "infeasible" to match the intended meaning.
  3. [Sec. 5.4, Table 5] The table header/row spells "NFBNET" but the method name is "NBFNET" (also in Sec. 4.1); please make the spelling consistent.
  4. [References] In the reference to Kipf & Welling, "Varitional graph auto-encoders" should be "Variational graph auto-encoders."
  5. [Sec. 5.1] The statement that "all GNN-based models utilize the same GNN backbone" is imprecise because NBFNet's message passing and readout differ from GraphSAGE; please clarify which components (tabular encoder, hidden sizes, GNN layers) are shared and which are architecture-specific.
  6. [Sec. 5.3] The sentence beginning "As per evaluation protocol" contains a stray "et al." before the parenthetical citations; please fix the citation formatting.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ContextGNN’s scores, fallback, and fusion are independently defined and the evaluation is against public benchmarks.

full rationale

The paper’s derivation chain is self-contained. The pair-wise local scores, the shallow-item two-tower fallback scores, and the fused final score in Eq. (2) are independent architectural definitions; the user-specific fusion offset MLP is a learned parameter, not a quantity constructed from the evaluation metric. The locality score in Eq. (1) is a dataset statistic computed from ground-truth links and is not defined in terms of ContextGNN’s outputs, so using it to motivate a fallback is not circular. The self-citations to RelBench, PyTorch Frame, and Relational Deep Learning are to public benchmarks and frameworks by overlapping authors, but those resources provide externally defined data and infrastructure rather than loading the central claim with an unverified premise; under the review rules, such citations are real evidence and do not raise the circularity score. The reported 20% average gain over NBFNet is potentially confounded by candidate-set coverage, since no NBFNet-plus-simple-fallback baseline is run, and Eq. (2) as printed makes the fallback branch unreachable because every item is in R; however, these are experimental-control and correctness concerns, not cases where a prediction reduces by construction to its inputs. Accordingly, no circular step is identified.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper makes no first-principles derivation. It introduces a hybrid architecture and supports it empirically. The ledger records the hand-set hyperparameters and the modeling assumptions that are load-bearing for the empirical claim. There are no invented physical or theoretical entities.

free parameters (6)
  • Learned model weights (user GNN, fusion MLP, shallow item embeddings) = optimized by cross-entropy on each training set
    These are the fitted parameters of the architecture, not derived from first principles; listed for completeness.
  • Subgraph depth k = 3 (implied, not explicitly reported per task)
    Defines the local candidate set and locality score; Sec. 3 states depth beyond 3 is infeasible. Chosen by scalability, not derived.
  • Hidden units = tuned from {32, 64, 128, 256, 512} per task
    Hyperparameter selected by validation performance; reported in Sec. 5.1.
  • Batch size = tuned from {256, 512, 1024} per task
    Hyperparameter selected by validation performance; reported in Sec. 5.1.
  • Learning rate = tuned from {0.001, 0.01}
    Hyperparameter selected by validation performance; reported in Sec. 5.1.
  • Number of sampled softmax classes C = up to approximately 1 million
    Chosen for GPU memory and training signal; described in Sec. 4.4.
assumptions (5)
  • domain assumption Sampling a k-hop subgraph and treating it as bidirectional preserves the temporal and behavioral information needed for pair-wise ranking.
    Sec. 4.1 steps 1 to 3 and the bidirectional transformation paragraph; this is the basis of the local candidate set.
  • domain assumption For items outside the subgraph, inner product of the user GNN representation with a shallow item embedding is a sufficient ranking signal.
    Sec. 4.2 and Eq. 2; this is the main fallback premise and the weakest assumption.
  • domain assumption Shallow item embeddings capture popularity, seasonal, demographic, and similarity signals as well as a full item-side GNN.
    Sec. 4.2 bullets; no proof, and Sec. 5.1 admits that deep item-side GNNs help on condition-sponsor-run.
  • ad hoc to paper An additive user-specific scalar offset from an MLP is sufficient to combine pair-wise and two-tower scores into one ranking.
    Eq. 2; the fusion form is chosen for simplicity and is not derived from an optimality criterion.
  • domain assumption The sampled softmax with ground-truth items always included in the class set gives a training signal that transfers to the reported ranking metrics.
    Sec. 4.4; the priority queue is a nonstandard sampling scheme whose bias is not analyzed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ContextGNN: Beyond Two-Tower Recommendation Systems." pith.science (2026). https://pith.science/paper/6NDFDNML

@misc{pith2026241119513,
  author       = {Pith},
  title        = {Pith review of: ContextGNN: Beyond Two-Tower Recommendation Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6NDFDNML}},
  note         = {Machine review of arXiv:2411.19513}
}
read the original abstract

Recommendation systems predominantly utilize two-tower architectures, which evaluate user-item rankings through the inner product of their respective embeddings. However, one key limitation of two-tower models is that they learn a pair-agnostic representation of users and items. In contrast, pair-wise representations either scale poorly due to their quadratic complexity or are too restrictive on the candidate pairs to rank. To address these issues, we introduce Context-based Graph Neural Networks (ContextGNNs), a novel deep learning architecture for link prediction in recommendation systems. The method employs a pair-wise representation technique for familiar items situated within a user's local subgraph, while leveraging two-tower representations to facilitate the recommendation of exploratory items. A final network then predicts how to fuse both pair-wise and two-tower recommendations into a single ranking of items. We demonstrate that ContextGNN is able to adapt to different data characteristics and outperforms existing methods, both traditional and GNN-based, on a diverse set of practical recommendation tasks, improving performance by 20% on average.

Figures

Figures reproduced from arXiv: 2411.19513 by the authors.

Figure 1
Figure 1. Overview of Context-based Graph Neural Networks. CONTEXTGNN utilizes a bidi￾rectional GNNθ to learn user h (2) v and user-specific item representations h (2) w within a user’s local subgraph. Its message passing scheme is enhanced by additionally propagating shallow item em￾beddings ww and seed user IDENTICATORθ representations. Afterwards, item scores are produced depending on whether an item situates within a user… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. RelGNN: Composite Message Passing for Relational Deep Learning

    cs.LG 2025-02 conditional novelty 6.0 of 10

    RelGNN passes messages along schema-derived atomic routes through junction tables and improves over heterogeneous GNN baselines on most RelBench tasks, with the largest gain on one trial-database regression task.

Reference graph

Works this paper leans on

60 extracted references · 52 canonical work pages · cited by 1 Pith paper

  1. [1]

    J. S. Baras and G. Theodorakopoulos. Path problems in networks. Synthesis Lectures on Communication Networks, 3 0 (1): 0 1--77, 2010

    work page 2010

  2. [2]

    X. Cai, C. Huang, L. Xia, and X. Ren. LightGCL : Simple yet effective graph contrastive learning for recommendation. In ICLR, 2023

    work page 2023

  3. [3]

    M. G. Campana and F. Delmastro. Recommender systems for online and mobile social networks: A survey. Online Social Networks and Media, 3: 0 75--97, 2017

    work page 2017

  4. [4]

    Fey and J

    M. Fey and J. E. Lenssen. Fast graph representation learning with PyTorch Geometric . In ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019

    2019

  5. [5]

    M. Fey, W. Hu, K. Huang, J. E. Lenssen, R. Ranjan, J. Robinson, R. Ying, J. You, and J. Leskovec. Relational deep learning: Graph representation learning on relational databases. In ICML, 2024

    work page 2024

  6. [6]

    C. Gao, X. He, D. Gan, X. Chen, F. Feng, Y. Li, T. S. Chua, and D. Jin. Neural multi-task recommendation from multi-behavior data. In ICDE, 2019

    work page 2019

  7. [7]

    Revisiting deep learning models for tabular data

    Yury Gorishniy, Ivan Rubachev, Valentin Khrulkov, and Artem Babenko. Revisiting deep learning models for tabular data. In NeurIPS, 2021

    work page 2021

  8. [8]

    Grover and J

    A. Grover and J. Leskovec. node2vec: Scalable feature learning for networks. In SIGKDD, 2016

    work page 2016

Show all 60 references
  1. [9]

    H. Guo, R. Tang, Y. Ye, Z. Li, and X. He. DeepFM : A factorization-machine based neural network for CTR prediction. In ICJAI, 2017

    2017

  2. [10]

    and Tat-Seng C

    Xiangnan H. and Tat-Seng C. Neural factorization machines for sparse predictive analytics. In SIGIR, 2017

    2017

  3. [11]

    Hamilton, Z

    W. Hamilton, Z. Ying, and J. Leskovec. Inductive representation learning on large graphs. In NIPS, 2017

    2017

  4. [12]

    He and J

    R. He and J. McAuley. Fusing similarity models with M arkov chains for sparse sequential recommendation. In ICDM, 2016

    2016

  5. [13]

    X. He, L. Liao, H. Zhang, L. Nie, X. Hu, and T. Chua. Neural collaborative filtering. In WWW, 2017

    2017

  6. [14]

    X. He, K. Deng, X. Wang, Y. Li, Y. Zhang, and M. Wang. LightGCN : Simplifying and powering graph convolution network for recommendation. In SIGIR, 2020

    2020

  7. [15]

    Z. He, W. Liu, W. Guo, J. Qin, Y. Zhang, Y. Hu, and R. Tang. A survey on user behavior modeling in recommender systems. In IJCAI, 2023

    2023

  8. [16]

    W. Hu, Y. Yuan, Z. Zhang, A. Nitta, K. Cao, V. Kocijan, J. Leskovec, and M. Fey. PyTorch Frame : A modular framework for multi-modal tabular learning. CoRR, abs/2404.00776, 2024

    2024 arXiv

  9. [17]

    Y. Hu, Y. Koren, and C. Volinsky. Collaborative filtering for implicit feedback datasets. In ICDM, 2008

    2008

  10. [18]

    Huang, F

    S. Huang, F. Poursafaei, J. Danovitch, M. Fey, W. Hu, E. Rossi, J. Leskovec, M. Bronstein, G. Rabusseau, and R. Rabbany. Temporal graph benchmark for machine learning on temporal graphs. NeurIPS, 2023

    2023

  11. [19]

    B. Jin, C. Gao, X. He, D. Jin, and Y. Li. Multi-behavior recommendation with graph convolutional networks. In SIGIR, 2020

    2020

  12. [20]

    Johnson, M

    J. Johnson, M. Douze, and H. J \'e gou. Billion-scale similarity search with GPUs . IEEE Transactions on Big Data, 7 0 (3): 0 535--547, 2019

    2019

  13. [21]

    Kang and J

    W. Kang and J. McAuley. Self-attentive sequential recommendation. In ICDM, 2018

    2018

  14. [22]

    G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T. Liu. LightGBM : A highly efficient gradient boosting decision tree. In NIPS, 2017

    2017

  15. [23]

    D. P. Kingma and J. L. Ba. Adam: A method for stochastic optimization. In ICLR, 2015

    2015

  16. [24]

    T. N. Kipf and M. Welling. Varitional graph auto-encoders. In NIPS Workshop on Bayesian Deep Learning, 2016

    2016

  17. [25]

    Koren, R

    Y. Koren, R. Bell, and C. Volinsky. Matrix factorization techniques for recommender systems. Computer, 42 0 (8): 0 30--37, 2009

    2009

  18. [26]

    J. Li, P. Ren, Z. Chen, Z. Ren, T. Lian, and J. Ma. Neural attentive session-based recommendation. New York, NY, USA, 2017

    2017

  19. [27]

    Y. Li, K. Liu, R. Satapathy, S. Wang, and E. Cambria. Recent developments in recommender systems: A survey. IEEE Computational Intelligence Magazine, 19: 0 78--95, 2024

    2024

  20. [28]

    Liang, R

    D. Liang, R. G. Krishnan, M. D. Hoffman, and T. Jebara. Variational autoencoders for collaborative filtering. In WWW, 2018

    2018

  21. [29]

    Q. Liu, Y. Zeng, R. Mokhosi, and H. Zhang. STAMP : Short-term attention/memory priority model for session-based recommendation. In SIGKDD, 2018

    2018

  22. [30]

    C. Ma, L. Ma, Y. Zhang, J. Sun, X. Liu, and M. Coates. Memory augmented graph neural networks for sequential recommendation. In AAAI, 2020

    2020

  23. [31]

    K. Mao, J. Zhu, X. Xiao, B. Lu, Z. Wang, and X. He. UltraGCN : Ultra simplification of graph convolutional networks for recommendation. In CIKM, 2021

    2021

  24. [32]

    Mnih and R

    A. Mnih and R. R. Salakhutdinov. Probabilistic matrix factorization. In NIPS, 2007

    2007

  25. [33]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito , M. Raison, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala. PyTorch : An imperative style, high-p...

    2019

  26. [34]

    Perozzi, R

    B. Perozzi, R. Al-Rfou, and S. Skiena. DeepWalk : Online learning of social representations. In SIGKDD, 2014

    2014

  27. [35]

    Rendle, C

    S. Rendle, C. Freudenthaler, Z. Gantner, and L. Schmidt-Thieme. BPR : Bayesian personalized ranking from implicit feedback. In UAI, 2009

    2009

  28. [36]

    Rendle, C

    S. Rendle, C. Freudenthaler, and L. Schmidt-Thieme. Factorizing personalized M arkov chains for next-basket recommendation. In WWW, 2010

    2010

  29. [37]

    Ricci, L

    F. Ricci, L. Rokach, and B. Shapira. Introduction to recommender systems handbook. In Recommender systems handbook. 2010

    2010

  30. [38]

    Robinson, R

    J. Robinson, R. Ranjan, W. Hu, K. Huang, J. Han, A. Dobles, M. Fey, J. E. Lenssen, Y. Yuan, Z. Zhang, X. He, and J. Leskovec. RelBench : A benchmark for deep learning on relational databases. In NeurIPS, 2024

    2024

  31. [39]

    Schlichtkrull, T

    M. Schlichtkrull, T. N. Kipf, P. Bloem, R. van den Berg, I. Titov, and M. Welling. Modeling relational data with graph convolutional networks. In The Semantic Web, 2018

    2018

  32. [40]

    F. Sun, J. Liu, J. Wu, C. Pei, X. Lin, W. Ou, and P. Jiang. BERT4Rec : Sequential recommendation with bidirectional encoder representations from transformer. In CIKM, 2019

    2019

  33. [41]

    K. K. Teru, E. G. Denis, and W. L. Hamilton. Inductive relation prediction by subgraph reasoning. In ICML, 2020

    2020

  34. [42]

    C. Wang, M. Zhang, W. Ma, Y. Liu, and Sh. Ma. Make it a chorus: knowledge-and time-aware item modeling for sequential recommendation. In SIGIR, 2020 a

    2020

  35. [43]

    J. Wang, K. Ding, L. Hong, H. Lui, and J. Caverlee. Next-item recommendation with sequential hypergraphs. In SIGIR, 2020 b

    2020

  36. [44]

    X. Wang, X. He, M. Wang, F. Feng, and T. Chua. Neural graph collaborative filtering. In SIGIR, 2019

    2019

  37. [45]

    J. Webber. Powering real-time recommendations with graph database technology. Technical report, Neo4J, 2021

    2021

  38. [46]

    J. Wu, X. Wang, F. Feng, X. He, L. Chen, J. Lian, and X. Xie. Self-supervised graph learning for recommendation. In SIGIR, 2021

    2021

  39. [47]

    L. Wu, Z. Zheng, Z. Qiu, H. Wang, H. Gu, T. Shen, C. Qin, C. Zhu, H. Zhu, Q. Liu, H. Xiong, and E. Chen. A survey on large language models for recommendation. CoRR, abs/2305.19860, 2023

    2023 arXiv

  40. [48]

    L. Xia, C. Huang, Y. Xu, P. Dai, B. Zhang, and L. Bo. Multiplex behavioral relation learning for recommendation via memory augmented transformer network. In SIGIR, 2020

    2020

  41. [49]

    L. Xia, C. Huang, Y. Xu, and J. Pei. Multi-behavior sequential recommendation with temporal graph transformer. In TKDE, 2022

    2022

  42. [50]

    Y. Yang, C. Huang, L. Xia, Y. Liang, Y. Yu, and C. Li. Multi-behavior hypergraph-enhanced transformer for sequential recommendation. In SIGKDD, 2022

    2022

  43. [51]

    J. You, R. Ying, and J. Leskovec. Position-aware graph neural networks. In ICML, 2019

    2019

  44. [52]

    J. You, J. M. Gomes-Selman, R. Ying, and J. Leskovec. Identity-aware graph neural networks. In AAAI, 2021

    2021

  45. [53]

    J. Yu, H. Yin, X. Xia, T. Chen, L. Cui, and Q. V. H. Nguyen. Are graph augmentations necessary? simple graph contrastive learning for recommendation. In SIGIR, 2022

    2022

  46. [54]

    H. Zeng, M. Zhang, Y. Xia, A. Srivastava, A. Malevich, R. Kannan, V. Prasanna, L. Jin, and R. Chen. Decoupling the depth and scope of graph neural networks. In NeurIPS, 2021

    2021

  47. [55]

    Zhang and Y

    M. Zhang and Y. Chen. Link prediction based on graph neural networks. In NIPS, 2018

    2018

  48. [56]

    Z. Zhu, Z. Zhang, L. P. Xhonneux, and J. Tang. N eural B ellmann- F ord networks: A general graph neural network framework for link prediction. In NeurIPS, 2021

    2021

  49. [57]

    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 gl...

  50. [58]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  51. [59]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  52. [60]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.