Pith. sign in

REVIEW 3 major objections 4 minor 32 references

Efficient Knowledge Tracing Leveraging Higher-Order Information in Integrated Graphs

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read DGAKT builds a small integrated subgraph for each student-exercise interaction and uses dual graph attention to achieve state-of-the-art knowledge tracing accuracy with about 60,000 parameters.

desk verdict The architecture is a reasonable synthesis, but the reported gains likely come from temporal leakage in subgraph construction, so the paper's central claims are unsupported until that is fixed. read the letter →

arxiv 2507.18668 v1 pith:FNRMH5WL submitted 2025-07-24 cs.LG cs.AI

Donghee Han , Daehee Kim , Minjun Lee , Daeyoung Roh , Keejun Han , Mun Yong Yi This is my paper · ORCID
classification cs.LGcs.AI
keywords knowledgetracinggraphneuralnetworkattentionsubgraphlearninghigher-orderinformationeducationaldataminingresourceefficiencyinductive
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

DGAKT claims that knowledge tracing can be made both more accurate and far cheaper by replacing the full student-exercise-KC graph with a small per-interaction subgraph. For each target student–exercise pair it pulls in the target student's recent exercises, the other students who attempted that exercise, and the exercise's knowledge concepts, then reads the merged graph with two attention layers. The paper reports the highest accuracy and AUC on three datasets — including AUC 0.8994 on ASSIST2017 versus 0.7985 for the best baseline — using only about 60,000 parameters. The result would matter because real online-learning platforms serve hundreds of thousands of students, and existing graph KT models scale quadratically with the number of exercises.

What carries the argument

The load-bearing machinery is the dual graph attention layer: a local edge-featured GAT layer that message-passes over subgraph edges, and a global attention layer that aggregates all node embeddings into a virtual subgraph node via one-hot node-type edges. Stacked twice, their outputs are concatenated into a subgraph embedding and a target node-pair embedding, and the final prediction mixes the two views with a weight parameter. The subgraph itself is built around a labeling trick that encodes roles as one-hot features.

What would settle it

Rebuild subgraphs using only interactions timestamped strictly before the target interaction, and exclude test-set edges from subgraph construction, then re-run the ASSIST2017 evaluation; if the reported AUC of 0.8994 falls toward the IGMC-KC baseline of 0.7985, the gains are leakage-driven.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that an integrated subgraph plus dual attention recovers higher-order paths that separate student-exercise and exercise-KC graphs miss. It introduces a labeling trick that assigns role-based one-hot features, stacks local edge-featured GAT layers over edges carrying timestamp, interaction-count and correctness features, and a global attention layer that pools through a virtual subgraph node; the two views are trained jointly with BCE losses plus an RMSE consistency term. The reported result is state-of-the-art ACC/AUC across EdNet, ASSIST2017 and Junyi, with ablations showing both attention mechanisms and all three edge features contribute, and type-split experiments suggesting robustness to unseen exercises.

Load-bearing premise

For each prediction, the model builds the subgraph from all students who attempted the target exercise, without restricting to responses that happened before the prediction time, so it assumes that future or test responses are legitimate input signals rather than leaked labels.

Editorial extensions

If this is right

  • Graph-based knowledge tracing can scale to datasets with over 100,000 students without materializing a full exercise-exercise adjacency matrix, since each prediction only needs its own subgraph.
  • Real-time inference becomes feasible on large platforms because every target interaction is an independent, parallelizable subgraph computation.
  • New exercises and knowledge concepts can be handled inductively through their roles in the subgraph, which the type-split experiments probe.
  • The local and global attention views provide a built-in interpretability signal: high-attention exercises and KCs in a subgraph highlight which prior knowledge matters for the target exercise.

Reading between the lines

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

  • The reported gains may be partly inflated by an untested temporal leak: subgraphs include all students who interacted with the target exercise regardless of whether those interactions occurred after the target timestamp, so a chronological re-split is the natural check.
  • The global attention layer over co-attempting students is effectively a collaborative-filtering signal; wiring a sequence model into the student node could recover the ordering the current subgraph discards.
  • The labeling trick fixes node roles as one-hot features, so a genuinely inductive deployment must decide how to assign labels when an unseen exercise brings previously unseen knowledge concepts into the subgraph.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes DGAKT, a graph neural network for knowledge tracing that constructs local subgraphs from an integrated student-exercise-KC graph and applies dual (local and global) graph attention layers to predict a student's correctness. The authors claim state-of-the-art accuracy and AUC on EdNet, ASSIST2017, and Junyi, plus superior resource efficiency with only 60k parameters. The experimental sections include overall comparisons, ablations, unseen-exercise tests, complexity analysis, and an interpretability case study.

Significance. If the empirical results were valid, the subgraph-based dual-attention design would be a useful contribution to large-scale knowledge tracing, particularly for inductive settings involving new exercises or KCs, and the small parameter count would be an attractive practical property. However, the evaluation as written appears to admit future and test-set interactions into the subgraphs, so the central performance claims are not supported. The efficiency claim is also not backed by actual timing or memory measurements. The paper's significance therefore hinges entirely on whether the leakage can be removed and the experiments redone.

major comments (3)
  1. [Section 2.2, Tables 2, 5, and 6] The subgraph construction described in Section 2.2 is not temporally causal. It selects 'the students who have interacted with the target exercise' and then includes 'all relationships between the selected nodes ... except the target interaction.' For a prediction at time t, this permits edges with timestamps after t, including edges from the validation and test portions of the chronological split. In the unseen-exercise experiments (Section 3.4, Tables 5-6), the test exercises are types not seen in training, so no training interaction with those exercises exists; the only 'students who have interacted with the target exercise' are test-set students. Because the local-attention edge features include the response (Section 2.3), DGAKT receives other students' actual responses to the same target exercise as input. The paper never states that only past interactions are used. This label leakage could fully explain the reported improvements (e.g., ASSIST2017 AUC 0.8994 vs 0.7985 in Table 2, and 0.9480 vs 0.8442 in Table 5), so the central performance claim is not supported.
  2. [Section 4.1, Table 7] The resource-efficiency claim is not backed by a meaningful comparison. In Table 7, DGAKT's time complexity is given as O(n·d^2 + n^2·d), but no definition of n for the subgraph setting is supplied, and it is not comparable to the sequence-length n used for DKT, AKT, and SAINT. The paper presents no wall-clock training time or peak-memory measurements; Figure 5 reports parameter counts only, which is not the same as computational cost or memory footprint. The abstract's 'new standard in resource efficiency' is therefore overstated as presented.
  3. [Section 3.3] The evaluation protocol is underspecified in a way that is load-bearing. The paper says splits are chronological but does not state that subgraphs for a test interaction are built exclusively from training-time edges, nor how the subsequence length interacts with the split. Without that statement, the leakage concern in the first major comment applies by default to every experiment in Table 2.
minor comments (4)
  1. [Section 3.4] The t-tests are reported only as p<0.01; with five repetitions, the absence of standard deviations or confidence intervals makes this hard to verify.
  2. [Section 3.3] The paper refers to a 'public repository' but gives no URL; the code and dataset split details are essential for verifying the temporal protocol.
  3. [Equation (2)] The min-max normalization is not specified clearly: it is unclear whether the minimum and maximum are computed per subgraph, per student sequence, or globally over the dataset.
  4. [Figure 7] The subgraph visualization labels are too small to read; the case study would benefit from larger fonts and an explicit description of the color scale and node types.

Circularity Check

1 steps flagged · score 6.0 of 10

DGAKT's subgraph construction admits future and test responses as edge features, so the headline performance gain is at least partly read off the target labels rather than derived independently.

  1. self definitional [Section 2.2 (Integrated subgraph construction) and Section 2.3 (Local graph attention); applied in the Section 3.4 unseen-case protocol.]
    "Student-exercise-KC subgraphs are constructed with the following three components: (1) the exercises in the target student's learning sequence, (2) the students who have interacted with the target exercise, and (3) the KCs connected to the exercises. In the subgraph, all relationships between the selected nodes are included except the target interaction. For local graph attention, the timestamp, number of previous interactions and the response from the student-exercise interactions are combined into 3-dimensional vectors and used as the initial edge features."

    The response variable r (correct/incorrect on the target exercise) is the prediction target, yet response values of the same kind are included as edge features in the input subgraph: 'the response from the student-exercise interactions ... used as the initial edge features.' Because subgraph selection imposes no timestamp filter ('all relationships ... included except the target interaction'), edges occurring after the target time, including held-out test responses, are available. The unseen-case split makes this exact: test exercise types are not exposed in training, so any 'students who have interacted with the target exercise' can only be test-set students whose response labels are ground truth.

full rationale

No load-bearing self-citation chain appears: the novelty claims are backed by comparisons to external baselines and by ablations, the complexity table is an accounting exercise rather than a first-principles derivation, and validation-based hyperparameter selection is standard practice. However, the construction defining the model input is self-definitional in the KT temporal sense. Section 2.2 selects 'the students who have interacted with the target exercise' and then includes 'all relationships ... except the target interaction,' with no temporal filter. Section 2.3 makes the response value an explicit edge feature. For a target interaction at time t, edges with timestamps after t, including held-out responses, can enter the subgraph. In the unseen-exercise protocol of Section 3.4, test exercise types are never exposed in training, so the only interactions with a target exercise are test-set interactions; hence other students' ground-truth responses to that exercise are fed in as edge features. The predicted response is therefore partially determined by the response labels it is meant to predict, undermining the independent content of Tables 2, 5, and 6. This is partial circularity rather than a self-citation chain, so the score is 6 rather than 8-10.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The model's central claims rest on several design assumptions and validation-tuned hyperparameters. The most fragile assumption is that subgraphs can include interactions across all times without breaking the temporal protocol; this is an unflagged assumption that directly affects the validity of the reported performance.

free parameters (5)
  • gamma (γ) = not reported
    Blending weight between local and global predictions (Eq. 9) and losses (Eq. 13); tuned per dataset, no values given.
  • lambda (λ) = not reported
    Weight of the consistency loss term in Eq. 13; tuned per dataset, no values given.
  • subsequence length = 8 for ASSIST2017, longer for EdNet/Junyi (Fig. 6)
    Non-overlapping subsequence length for subgraph construction; chosen per dataset based on validation performance.
  • number of layers L = 2
    Stated in Section 2.4; no search reported.
  • cap of previous interactions = 128
    Edge feature '# previous interactions' is capped at 128, then min-max normalized (Section 2.3).
assumptions (4)
  • domain assumption The labeling trick from IGMC (Zhang and Chen, 2020) provides sufficient node features: labeling target student 0, target exercise 1, neighboring students 2, neighboring exercises 3, and KCs 4/5, with one-hot vectors, is enough to represent the subgraph structure for prediction.
    Section 2.2 relies on this without theoretical justification; if these labels lose information about node identity, the model cannot generalize to unseen nodes.
  • domain assumption An integrated student-exercise-KC graph, including only the target student's sequence, students who interacted with the target exercise, and KCs connected to those exercises, captures all relevant higher-order information for the target interaction.
    Section 2.2 restricts the subgraph; any relevant signal from outside this neighborhood is inaccessible.
  • domain assumption The global attention virtual node with unidirectional edges aggregates node importance without oversmoothing and encodes higher-order information.
    Section 2.3 states this design 'can capture signals that are often attenuated by traditional multi-hop message passing'; it is an inductive bias, not proven.
  • domain assumption Chronological splitting of interactions into train/validation/test is a valid evaluation protocol for KT.
    Section 3.3 assumes the split prevents leakage, but subgraph construction may violate this by including future or test interactions without temporal filtering.
invented entities (1)
  • Virtual subgraph node (global attention node)
    purpose: Connected to all subgraph nodes with unidirectional edges; aggregates global attention and produces the subgraph embedding h_g.
    A modeling construct introduced for global pooling; no external, falsifiable prediction is associated with it. Ablations show it helps on the same datasets used for training, which is not independent evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Efficient Knowledge Tracing Leveraging Higher-Order Information in Integrated Graphs." pith.science (2026). https://pith.science/paper/FNRMH5WL

@misc{pith2026250718668,
  author       = {Pith},
  title        = {Pith review of: Efficient Knowledge Tracing Leveraging Higher-Order Information in Integrated Graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FNRMH5WL}},
  note         = {Machine review of arXiv:2507.18668}
}
read the original abstract

The rise of online learning has led to the development of various knowledge tracing (KT) methods. However, existing methods have overlooked the problem of increasing computational cost when utilizing large graphs and long learning sequences. To address this issue, we introduce Dual Graph Attention-based Knowledge Tracing (DGAKT), a graph neural network model designed to leverage high-order information from subgraphs representing student-exercise-KC relationships. DGAKT incorporates a subgraph-based approach to enhance computational efficiency. By processing only relevant subgraphs for each target interaction, DGAKT significantly reduces memory and computational requirements compared to full global graph models. Extensive experimental results demonstrate that DGAKT not only outperforms existing KT models but also sets a new standard in resource efficiency, addressing a critical need that has been largely overlooked by prior KT approaches.

Figures

Figures reproduced from arXiv: 2507.18668 by the authors.

Figure 1
Figure 1. Illustrative example of a Student-Exercise-KC graph and indirect paths between target student and exercise. Several paths that [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The framework of the proposed model, DGAKT. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Ablation study by edge features. indicate that all of the proposed variants showed inferior per￾formance compared to the original model. Unseen cases We further assess the ability of the proposed model to achieve robust predictive performance in unseen cases (RQ4). In on￾line education services, new courses, topics, and problems are often introduced. To verify that our model can infer robustly to these changes, we e… view at source ↗
Figures from the paper (2 more)
Figure 6
Figure 6. Figure 6: Performance changes with subsequence length [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 5
Figure 5. Figure 5: Comparing the number of parameters of KT models. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

  1. [1]

    Deep graph memory networks for forgetting- robust knowledge tracing

    [Abdelrahman and Wang, 2022] Ghodai Abdelrahman and Qing Wang. Deep graph memory networks for forgetting- robust knowledge tracing. IEEE Transactions on Knowl- edge and Data Engineering, pages 1–13,

    work page 2022

  2. [5]

    Higher-order clustering and pooling for graph neural networks

    [Duval and Malliaros, 2022] Alexandre Duval and Fragkiskos Malliaros. Higher-order clustering and pooling for graph neural networks. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management , CIKM ’22, page 426–435, New York, NY , USA,

    work page 2022

  3. [7]

    [Gao et al., 2023] Chen Gao, Yu Zheng, Nian Li, Yinfeng Li, Yingrong Qin, Jinghua Piao, Yuhan Quan, Jianxin Chang, Depeng Jin, Xiangnan He, and Yong Li

    Association for Computing Machinery. [Gao et al., 2023] Chen Gao, Yu Zheng, Nian Li, Yinfeng Li, Yingrong Qin, Jinghua Piao, Yuhan Quan, Jianxin Chang, Depeng Jin, Xiangnan He, and Yong Li. A survey of graph neural networks for recommender systems: Challenges, methods, and directions. ACM Trans. Recomm. Syst., 1(1), mar

    work page 2023

  4. [8]

    [Ghosh et al., 2020] Aritra Ghosh, Neil Heffernan, and An- drew S. Lan. Context-aware attentive knowledge tracing. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery &; Data Mining , KDD ’20, page 2330–2339, New York, NY , USA,

    work page 2020

  5. [9]

    [Huang et al., 2021] Chao Huang, Huance Xu, Yong Xu, Peng Dai, Lianghao Xia, Mengyin Lu, Liefeng Bo, Hao Xing, Xiaoping Lai, and Yanfang Ye

    Association for Computing Machinery. [Huang et al., 2021] Chao Huang, Huance Xu, Yong Xu, Peng Dai, Lianghao Xia, Mengyin Lu, Liefeng Bo, Hao Xing, Xiaoping Lai, and Yanfang Ye. Knowledge-aware coupled graph neural network for social recommendation. Proceedings of the AAAI Conference on Artificial Intelli- gence, 35(5):4115–4122, May

    work page 2021

  6. [11]

    Kingma and Jimmy Ba

    [Kingma and Ba, 2015] Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In Yoshua Bengio and Yann LeCun, editors, International Confer- ence on Learning Representations,

    work page 2015

  7. [13]

    Data- efficient fine-tuning for llm-based recommendation

    [Lin et al., 2024] Xinyu Lin, Wenjie Wang, Yongqi Li, Shuo Yang, Fuli Feng, Yinwei Wei, and Tat-Seng Chua. Data- efficient fine-tuning for llm-based recommendation. In Proceedings of the 47th International ACM SIGIR Con- ference on Research and Development in Information Re- trieval, pages 365–374,

    work page 2024

  8. [14]

    Improving knowledge tracing via pre-training question embeddings

    [Liu et al., 2020] Yunfei Liu, Yang Yang, Xianyu Chen, Jian Shen, Haifeng Zhang, and Yong Yu. Improving knowledge tracing via pre-training question embeddings. In Proceed- ings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI’20,

    work page 2020

Show all 32 references
  1. [15]

    pykt: A python library to benchmark deep learning based knowl- edge tracing models

    [Liu et al., 2022] Zitao Liu, Qiongqiong Liu, Jiahao Chen, Shuyan Huang, Jiliang Tang, and Weiqi Luo. pykt: A python library to benchmark deep learning based knowl- edge tracing models. In S. Koyejo, S. Mohamed, A. Agar- wal, D. Belgrave, K. Cho, and A. Oh, editors,Advances in...

    2022

  2. [16]

    Enhancing deep knowledge tracing with auxiliary tasks

    [Liu et al., 2023] Zitao Liu, Qiongqiong Liu, Jiahao Chen, Shuyan Huang, Boyu Gao, Weiqi Luo, and Jian Weng. Enhancing deep knowledge tracing with auxiliary tasks. In Proceedings of the ACM Web Conference 2023, WWW ’23, page 4178–4187, New York, NY , USA,

    2023

  3. [17]

    [Nakagawa et al., 2019] Hiromi Nakagawa, Yusuke Iwa- sawa, and Yutaka Matsuo

    Associ- ation for Computing Machinery. [Nakagawa et al., 2019] Hiromi Nakagawa, Yusuke Iwa- sawa, and Yutaka Matsuo. Graph-based knowledge trac- ing: Modeling student proficiency using graph neural net- work. In 2019 IEEE/WIC/ACM International Conference on Web Intelligence (W...

    2019

  4. [19]

    Pytorch: An imperative style, high- performance deep learning library

    [Paszke et al., 2019] Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurth...

    2019

  5. [20]

    Deep knowledge trac- ing

    [Piech et al., 2015] Chris Piech, Jonathan Bassen, Jonathan Huang, Surya Ganguli, Mehran Sahami, Leonidas J Guibas, and Jascha Sohl-Dickstein. Deep knowledge trac- ing. Advances in neural information processing systems , 28,

    2015

  6. [21]

    Saint+: Integrating temporal features for ednet correctness prediction

    [Shin et al., 2021] Dongmin Shin, Yugeun Shim, Hangyeol Yu, Seewoo Lee, Byungsoo Kim, and Youngduck Choi. Saint+: Integrating temporal features for ednet correctness prediction. In LAK21: 11th International Learning Ana- lytics and Knowledge Conference, LAK21, page 490–496, Ne...

    2021

  7. [22]

    [Sun et al., 2022] Jianwen Sun, Rui Zou, Ruxia Liang, Lu Gao, Sannyuya Liu, Qing Li, Kai Zhang, and Lulu Jiang

    Association for Computing Machinery. [Sun et al., 2022] Jianwen Sun, Rui Zou, Ruxia Liang, Lu Gao, Sannyuya Liu, Qing Li, Kai Zhang, and Lulu Jiang. Ensemble knowledge tracing: Modeling interac- tions in learning process. Expert Systems with Applica- tions, 207:117680,

    2022

  8. [23]

    Introducing prob- lem schema with hierarchical exercise graph for knowl- edge tracing

    [Tong et al., 2022] Hanshuang Tong, Zhen Wang, Yun Zhou, Shiwei Tong, Wenyuan Han, and Qi Liu. Introducing prob- lem schema with hierarchical exercise graph for knowl- edge tracing. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in In...

    2022

  9. [24]

    Association for Computing Ma- chinery. [Touvron et al., 2023] Hugo Touvron, Thibaut Lavril, Gau- tier Izacard, Xavier Martinet, Marie-Anne Lachaux, Tim- oth´ee Lacroix, Baptiste Rozi`ere, Naman Goyal, Eric Ham- bro, Faisal Azhar, et al. Llama: Open and efficient founda- tion l...

    2023 arXiv

  10. [25]

    Graph attention networks

    [Veliˇckovi´c et al., 2018] Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Li `o, and Yoshua Bengio. Graph attention networks. In Interna- tional Conference on Learning Representations,

    2018

  11. [27]

    Egat: Edge-featured graph attention network

    [Wang et al., 2021] Ziming Wang, Jun Chen, and Haopeng Chen. Egat: Edge-featured graph attention network. In International Conference on Artificial Neural Networks , pages 253–264. Springer,

    2021

  12. [28]

    Self- supervised heterogeneous hypergraph network for knowl- edge tracing

    [Wu and Ling, 2023] Tangjie Wu and Qiang Ling. Self- supervised heterogeneous hypergraph network for knowl- edge tracing. Information Sciences, 624:200–216,

    2023

  13. [29]

    Gikt: a graph-based interaction model for knowledge tracing

    [Yang et al., 2020] Yang Yang, Jian Shen, Yanru Qu, Yun- fei Liu, Kerong Wang, Yaoming Zhu, Weinan Zhang, and Yong Yu. Gikt: a graph-based interaction model for knowledge tracing. In Joint European Conference on Ma- chine Learning and Knowledge Discovery in Databases , pages 2...

    2020

  14. [30]

    In- ductive matrix completion based on graph neural net- works

    [Zhang and Chen, 2020] Muhan Zhang and Yixin Chen. In- ductive matrix completion based on graph neural net- works. In International Conference on Learning Repre- sentations,

    2020

  15. [31]

    Dynamic key-value memory networks for knowledge tracing

    [Zhang et al., 2017] Jiani Zhang, Xingjian Shi, Irwin King, and Dit-Yan Yeung. Dynamic key-value memory networks for knowledge tracing. In Proceedings of the 26th inter- national conference on World Wide Web, pages 765–774,

    2017

  16. [2015]

    Yu, and Lifang He

    [Li et al., 2023] Jianxin Li, Hao Peng, Yuwei Cao, Yingtong Dou, Hekai Zhang, Philip S. Yu, and Lifang He. Higher- order attribute-enhancing heterogeneous graph neural net- works. IEEE Transactions on Knowledge and Data Engi- neering, 35(1):560–574,

    2023

  17. [2017]

    Inhomogeneous Social Rec- ommendation with Hypergraph Convolutional Networks

    [Zhu et al., 2022] Zirui Zhu, Chen Gao, Xu Chen, Nian Li, Depeng Jin, and Yong Li. Inhomogeneous Social Rec- ommendation with Hypergraph Convolutional Networks. 2022 IEEE 38th International Conference on Data Engi- neering (ICDE), 2022

    2022

  18. [2018]

    Deep graph li- brary: A graph-centric, highly-performant package for graph neural networks,

    [Wang et al., 2019] Minjie Wang, Da Zheng, Zihao Ye, Quan Gan, Mufei Li, Xiang Song, Jinjing Zhou, Chao Ma, Lingfan Yu, Yu Gai, Tianjun Xiao, Tong He, George Karypis, Jinyang Li, and Zheng Zhang. Deep graph li- brary: A graph-centric, highly-performant package for graph neural...

    2019

  19. [2019]

    A self-attentive model for knowledge tracing

    [Pandey and Karypis, 2019] Shalini Pandey and George Karypis. A self-attentive model for knowledge tracing. In 12th International Conference on Educational Data Min- ing, EDM 2019, pages 384–389. International Educational Data Mining Society,

    2019

  20. [2020]

    [Cui et al., 2024] Chaoran Cui, Yumo Yao, Chunyun Zhang, Hebo Ma, Yuling Ma, Zhaochun Ren, Chen Zhang, and James Ko

    Springer International Publishing. [Cui et al., 2024] Chaoran Cui, Yumo Yao, Chunyun Zhang, Hebo Ma, Yuling Ma, Zhaochun Ren, Chen Zhang, and James Ko. Dgekt: A dual graph ensemble learning method for knowledge tracing. ACM Trans. Inf. Syst. , 42(3), jan

    2024

  21. [2021]

    Large language models meet collaborative filter- ing: An efficient all-round llm-based recommender sys- tem

    [Kim et al., 2024] Sein Kim, Hongseok Kang, Seungyoon Choi, Donghyun Kim, Minchul Yang, and Chanyoung Park. Large language models meet collaborative filter- ing: An efficient all-round llm-based recommender sys- tem. arXiv preprint arXiv:2404.11343,

    2024 arXiv

  22. [2022]

    [Gao et al., 2022] Chen Gao, Xiang Wang, Xiangnan He, and Yong Li

    Association for Computing Machinery. [Gao et al., 2022] Chen Gao, Xiang Wang, Xiangnan He, and Yong Li. Graph neural networks for recommender system. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining, WSDM ’22, page 1623–1625, New York,...

    2022

  23. [2023]

    [Barron, 2017] Jonathan T. Barron. Continuously differen- tiable exponential linear units. CoRR, abs/1704.07483,

    2017 arXiv

  24. [2024]

    Online knowledge level tracking with data- driven student models and collaborative filtering

    [Cully and Demiris, 2020] Antoine Cully and Yiannis Demiris. Online knowledge level tracking with data- driven student models and collaborative filtering. IEEE Transactions on Knowledge and Data Engineering , 32(10):2000–2013,

    2020

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.