Uncertainty-aware Generative Learning Path Recommendation with Cognition-Adaptive Diffusion

Baiyang Chen; Hang Liang; Xiangrui Xiong; Yanli Lee; Zifei Pan

arxiv: 2604.14613 · v1 · submitted 2026-04-16 · 💻 cs.IR · cs.AI

Uncertainty-aware Generative Learning Path Recommendation with Cognition-Adaptive Diffusion

Xiangrui Xiong , Hang Liang , Baiyang Chen , Zifei Pan , Yanli Lee This is my paper

Pith reviewed 2026-05-10 10:09 UTC · model grok-4.3

classification 💻 cs.IR cs.AI

keywords learning path recommendationuncertainty modelingdiffusion modelscognitive statespersonalized educationgenerative recommendationGaussian LSTMgoal-oriented encoding

0 comments

The pith

U-GLAD models cognitive uncertainty as Gaussian distributions and uses diffusion to generate goal-aligned next concepts in learning paths.

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

Learning path recommendation often overlooks uncertainty in learner interactions such as lucky guesses or slips, and fails to adapt paths to varied goals. The paper introduces U-GLAD to represent a learner's cognitive state as a probability distribution via Gaussian LSTM, then applies a goal-oriented encoder with multi-head attention to create tailored concept embeddings. A generative diffusion model predicts the latent representation of the next optimal concept instead of using discriminative ranking. Evaluations on three public datasets show improved performance over baselines, with better handling of uncertainty and more stable, goal-driven paths.

Core claim

By modeling cognitive states as Gaussian distributions with an LSTM, employing a goal-oriented concept encoder that uses multi-head attention and objective-specific transformations, and applying a generative diffusion process to produce the latent representation of the next optimal concept, the framework generates personalized learning paths that account for interaction uncertainty and align with individual goals, outperforming traditional approaches on public datasets.

What carries the argument

The cognition-adaptive diffusion model that generates the latent representation of the next optimal concept, driven by Gaussian LSTM uncertainty modeling and multi-head attention for goal alignment.

If this is right

Recommendations gain stability by explicitly capturing uncertainty in historical interactions.
Concept embeddings become uniquely aligned with each learner's specific objectives through dynamic transformations.
Performance exceeds representative baselines on three public educational datasets.
The shift to generative prediction replaces ranking-based selection for more personalized outputs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Similar uncertainty distributions could improve recommendation stability in other domains with noisy user signals, such as content streaming.
Real-time path adjustments might become possible if the diffusion process scales to live interaction streams.
Representing knowledge as distributions rather than fixed values may reduce systematic bias in educational systems trained on imperfect data.

Load-bearing premise

That representing cognitive states as Gaussian distributions and generating next concepts via diffusion will reliably yield optimal goal-aligned paths without overfitting to dataset noise.

What would settle it

On a new held-out dataset, generated paths produce no measurable gain in learner completion rates or goal achievement compared with non-uncertainty baselines, or the modeled uncertainty fails to correlate with observed slip and guess patterns.

Figures

Figures reproduced from arXiv: 2604.14613 by Baiyang Chen, Hang Liang, Xiangrui Xiong, Yanli Lee, Zifei Pan.

**Figure 2.** Figure 2: Visualization of ablation studies across three datasets. We evaluate the contri [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗

**Figure 3.** Figure 3: Cognitive state instability analysis. The heatmaps visualize the state instability [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗

**Figure 4.** Figure 4: Impact of the number of diffusion model reverse process iterations [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗

read the original abstract

Learning Path Recommendation (LPR) is critical for personalized education, yet current methods often fail to account for historical interaction uncertainty (e.g., lucky guesses or accidental slips) and lack adaptability to diverse learning goals. We propose U-GLAD (Uncertainty-aware Generative Learning Path Recommendation with Cognition-Adaptive Diffusion). To address representation bias, the framework models cognitive states as probability distributions, capturing the learner's underlying true state via a Gaussian LSTM. To ensure highly personalized recommendation, a goal-oriented concept encoder utilizes multi-head attention and objective-specific transformations to dynamically align concept semantics with individual learning goals, generating uniquely tailored embeddings. Unlike traditional discriminative ranking approaches, our model employs a generative diffusion model to predict the latent representation of the next optimal concept. Extensive evaluations on three public datasets demonstrate that U-GLAD significantly outperforms representative baselines. Further analyses confirm its superior capability in perceiving interaction uncertainty and providing stable, goal-driven recommendation paths.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

U-GLAD pairs a Gaussian LSTM for uncertain cognitive states with diffusion to generate goal-aligned learning paths, but the outperformance claims lack the numbers and controls needed to confirm the gains come from those choices.

read the letter

The key takeaway is that this paper proposes a generative approach to learning path recommendation using diffusion models on top of uncertainty-aware cognitive state modeling, but the reported gains over baselines rest on claims without visible supporting numbers or controls. What stands out as new is the shift from typical ranking-based methods to generating latent representations of the next concept via diffusion, combined with a Gaussian LSTM to represent cognitive states probabilistically and multi-head attention for goal-specific embeddings. That integration for handling interaction uncertainty and adapting to different objectives looks like a fresh application in the LPR literature. The framework makes sense on paper for addressing lucky guesses and slips by treating states as distributions rather than point estimates. The goal-oriented encoder also seems practical for making recommendations more tailored. The main weakness is the lack of concrete evidence. The abstract mentions significant outperformance on three datasets and superior uncertainty handling from further analyses, yet no specific metrics, statistical tests, or ablation studies are referenced here. Without those, or checks like varying noise levels in interaction histories, it's difficult to link the improvements directly to the Gaussian and diffusion components rather than other design choices or hyperparameter effects. The stress-test concern about robustness to dataset noise appears valid based on what's described. This work would interest people building or studying personalized education platforms and sequential recommendation systems. Readers looking for ideas on incorporating generative models into edtech could find the architecture useful to build on. It deserves peer review because the core ideas address real limitations in current LPR methods and the approach is technically coherent, even if the experiments need strengthening. I would send it to referees with notes to verify the empirical claims and add robustness tests.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes U-GLAD for learning path recommendation (LPR). It models learner cognitive states as Gaussian distributions via LSTM to capture interaction uncertainties (e.g., lucky guesses or slips), uses multi-head attention with objective-specific transformations for goal-oriented concept embeddings, and employs a generative diffusion process to produce the latent representation of the next optimal concept. The central claims are that this yields significant outperformance over representative baselines on three public datasets plus superior uncertainty perception and stable, goal-driven paths.

Significance. If the empirical results and ablations hold, the work could meaningfully advance personalized education systems by shifting from discriminative ranking to generative diffusion conditioned on probabilistic cognitive states. This addresses a recognized gap in handling noisy learner data and goal adaptability; the explicit modeling of uncertainty as Gaussians and the diffusion-based generation represent a coherent technical departure from prior LPR methods.

major comments (2)

[Abstract and §5 (Experimental Evaluation)] Abstract and §5 (Experimental Evaluation): the central claim of 'significant outperformance' and 'superior capability in perceiving interaction uncertainty' is asserted without any reported metrics, error bars, run-to-run variance, ablation results on the Gaussian LSTM component, or statistical significance tests. This directly undermines verification that observed gains arise from the uncertainty modeling and diffusion step rather than dataset-specific artifacts or hyperparameter choices.
[§3.3 (Diffusion Model)] §3.3 (Diffusion Model): the description of the cognition-adaptive diffusion process does not specify the conditioning mechanism (e.g., how the Gaussian cognitive state parameters or goal-specific embeddings are injected into the reverse diffusion steps or noise schedule). Without these details or an equation showing the conditioned denoising objective, it is impossible to confirm that the generative step reliably produces goal-aligned paths that are causally linked to the uncertainty modeling.

minor comments (1)

[§3 (Method)] Notation for the Gaussian LSTM variance parameters and the multi-head attention weights is introduced without a consolidated table of symbols, making it harder to trace free parameters across the model description.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and positive evaluation of the significance of our work. We address the major comments point by point below, agreeing that additional details will strengthen the manuscript. Revisions have been made accordingly.

read point-by-point responses

Referee: [Abstract and §5 (Experimental Evaluation)] Abstract and §5 (Experimental Evaluation): the central claim of 'significant outperformance' and 'superior capability in perceiving interaction uncertainty' is asserted without any reported metrics, error bars, run-to-run variance, ablation results on the Gaussian LSTM component, or statistical significance tests. This directly undermines verification that observed gains arise from the uncertainty modeling and diffusion step rather than dataset-specific artifacts or hyperparameter choices.

Authors: We acknowledge the validity of this concern. Although the manuscript presents comparative results on three public datasets in §5, we agree that the reporting can be improved for better verifiability. In the revised version, we will augment the experimental evaluation with error bars (standard deviations over 5 runs), explicit run-to-run variance, a specific ablation study on the Gaussian LSTM (comparing to deterministic LSTM), and statistical significance tests (e.g., t-tests with p-values reported). This will help confirm that the gains are attributable to the proposed components rather than other factors. revision: yes
Referee: [§3.3 (Diffusion Model)] §3.3 (Diffusion Model): the description of the cognition-adaptive diffusion process does not specify the conditioning mechanism (e.g., how the Gaussian cognitive state parameters or goal-specific embeddings are injected into the reverse diffusion steps or noise schedule). Without these details or an equation showing the conditioned denoising objective, it is impossible to confirm that the generative step reliably produces goal-aligned paths that are causally linked to the uncertainty modeling.

Authors: We agree that the description in §3.3 requires more explicit technical details to allow full reproduction and understanding. The current text provides an overview but omits the precise injection method. In the revision, we will expand this section to detail the conditioning mechanism, including how the Gaussian cognitive state parameters and goal-specific embeddings are incorporated into the reverse diffusion steps. We will also introduce a new equation for the conditioned denoising objective that incorporates these elements, thereby clarifying the causal link to uncertainty modeling and goal alignment. revision: yes

Circularity Check

0 steps flagged

No significant circularity; model components and claims rest on external dataset evaluations.

full rationale

The paper defines U-GLAD via Gaussian LSTM for cognitive state distributions, multi-head attention for goal-aligned concept embeddings, and a diffusion process to generate next-concept latents; these are architectural choices trained on data rather than self-definitions. The central claims (outperformance on three public datasets plus superior uncertainty perception) are asserted via empirical results and 'further analyses,' not by renaming fitted parameters as predictions or importing uniqueness via self-citation. No equations or steps reduce the output to the input by construction; the generative step produces novel latents evaluated against held-out data. This is the common honest case of a self-contained empirical proposal with no load-bearing circular reductions.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The central claims rest on standard ML assumptions plus several unstated modeling choices; no machine-checked proofs or parameter-free derivations are present.

free parameters (3)

Gaussian LSTM variance parameters
Fitted to capture interaction uncertainty; values not specified but required for distribution modeling.
Diffusion model noise schedule and steps
Chosen to generate next-concept latents; typical free parameters in diffusion setups.
Multi-head attention and objective-specific transformation weights
Learned to align concept embeddings with learning goals.

axioms (2)

domain assumption Learner interactions contain measurable uncertainty that can be modeled as Gaussian noise around a true cognitive state
Invoked to justify the Gaussian LSTM representation of cognitive states.
domain assumption Diffusion processes can generate optimal next-concept representations from learned latents
Central to replacing discriminative ranking with generative prediction.

pith-pipeline@v0.9.0 · 5462 in / 1478 out tokens · 39843 ms · 2026-05-10T10:09:36.839279+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

[1]

A. Sun, X. Chen, Online education and its effective practice: A research review, J. Inf. Technol. Educ. Res. 15 (2016) 157–190

work page 2016
[2]

M. D. B. Castro, G. M. Tumibay, A literature review: efficacy of online learning courses for higher education institution using meta-analysis, Educ. Inf. Technol. 26 (2021) 1367–1385. 16

work page 2021
[3]

A. D. Dumford, A. L. Miller, Online learning in higher education: ex- ploring advantages and disadvantages for engagement, J. Comput. High. Educ. 30 (2018) 452–465

work page 2018
[4]

M. L. Bernacki, M. J. Greene, N. G. Lobczowski, A systematic review of research on personalized learning: Personalized by whom, to what, how, and for what purpose (s)?, Educ. Psychol. Rev. 33 (2021) 1675–1715

work page 2021
[5]

O. O. Ayeni, N. M. Al Hamad, O. N. Chisom, B. Osawaru, O. E. Adewusi, AI in education: A review of personalized learning and ed- ucational technology, GSC Adv. Res. Rev. 18 (2024) 261–271

work page 2024
[6]

A. H. Nabizadeh, J. P. Leal, H. N. Rafsanjani, R. R. Shah, Learning path personalization and recommendation methods: A survey of the state-of-the-art, Expert Syst. Appl. 159 (2020) 113596

work page 2020
[7]

N. W. Rahayu, R. Ferdiana, S. S. Kusumawardani, A systematic review of learning path recommender systems, Educ. Inf. Technol. 28 (2023) 7437–7460

work page 2023
[8]

D. Shi, T. Wang, H. Xing, H. Xu, A learning path recommen- dation model based on a multidimensional knowledge graph frame- work for e-learning, Knowledge-Based Systems 195 (2020) 105618. doi:10.1016/j.knosys.2020.105618

work page doi:10.1016/j.knosys.2020.105618 2020
[9]

Huang, Q

Z. Huang, Q. Liu, C. Zhai, Y. Yin, E. Chen, W. Gao, G. Hu, Exploring multi-objective exercise recommendations in online education systems, in: In Proc. ACM CIKM Conf., 2019, pp. 1261–1270

work page 2019
[10]

X. Chen, J. Shen, W. Xia, J. Jin, Y. Song, W. Zhang, W. Liu, M. Zhu, R. Tang, K. Dong, Set-to-sequence ranking-based concept-aware learn- ing path recommendation, in: In Proc. AAAI Conf., Vol. 37, 2023, pp. 5027–5035

work page 2023
[11]

Q. Li, W. Xia, L. Yin, J. Shen, R. Rui, W. Zhang, X. Chen, R. Tang, Y. Yu, Graph enhanced hierarchical reinforcement learning for goal- oriented learning path recommendation, in: In Proc. ACM CIKM Conf., 2023, pp. 1318–1327. 17

work page 2023
[12]

Zhang, S

H. Zhang, S. Shen, B. Xu, Z. Huang, J. Wu, J. Sha, S. Wang, Item- difficulty-aware learning path recommendation: From a real walking perspective, in: In Proc. ACM SIGKDD Conf., 2024, pp. 4167–4178

work page 2024
[13]

G. Luo, H. Gu, X. Dong, D. Zhou, HA-LPR: A highly adaptive learning path recommendation, Educ. Inf. Technol. 30 (2025) 14597–14627

work page 2025
[14]

X. Yu, S. Yang, Z. Wang, S. Song, H. Ma, Z. Cao, X. Zhang, LIGHT: Enhancinglearningpathrecommendationviaknowledgetopology-aware sequence optimization, in: In Proc. ACM SIGIR Conf., 2025, pp. 306– 315

work page 2025
[15]

Cheng, Z

X. Cheng, Z. Zhang, J. Wang, L. Fang, C. He, Q. Guan, S. Pan, W. Luo, GraphRAG-induced dual knowledge structure graphs for personalized learning path recommendation, in: In Proc. AAAI Conf., Vol. 40, 2026, pp. 14610–14620

work page 2026
[16]

Cheng, H

W. Cheng, H. Du, C. Li, E. Ni, L. Tan, T. Xu, Y. Ni, Uncertainty- aware knowledge tracing, in: In Proc. AAAI Conf., Vol. 39, 2025, pp. 27905–27913

work page 2025
[17]

Order Matters: Sequence to sequence for sets

O. Vinyals, S. Bengio, M. Kudlur, Order matters: Sequence to sequence for sets, arXiv preprint arXiv:1511.06391 (2015)

work page Pith review arXiv 2015
[18]

Z. Li, A. Sun, C. Li, DiffuRec: A diffusion model for sequential recom- mendation, ACM Trans. Inf. Syst. 42 (2023) 1–28

work page 2023
[19]

J. Ho, A. Jain, P. Abbeel, Denoising diffusion probabilistic models, Ad- vances in neural information processing systems 33 (2020) 6840–6851

work page 2020
[20]

Birjali, A

M. Birjali, A. Beni-Hssane, M. Erritali, A novel adaptive E-learning model based on Big Data by using competence-based knowledge and social learner activities, Applied Soft Computing 69 (2018) 14–32

work page 2018
[21]

M. Zare, C. Pahl, H. Rahnama, M. Nilashi, A. Mardani, O. Ibrahim, H. Ahmadi, Multi-criteria decision making approach in E-learning: A systematic review and classification, Applied Soft Computing 45 (2016) 108–128

work page 2016
[22]

Dwivedi, V

P. Dwivedi, V. Kant, K. K. Bharadwaj, Learning path recommendation based on modified variable length genetic algorithm, Educ. Inf. Technol. 23 (2018) 819–836. 18

work page 2018
[23]

Q. Liu, S. Tong, C. Liu, H. Zhao, E. Chen, H. Ma, S. Wang, Exploiting cognitive structure for adaptive learning, in: In Proc. ACM SIGKDD Conf., 2019, pp. 627–635

work page 2019
[24]

Gavrilovic, T

N. Gavrilovic, T. Sibalija, D. Domazet, Design and implementation of discrete Jaya and discrete PSO algorithms for automatic collaborative learning group composition in an e-learning system, Applied Soft Com- puting 129 (2022) 109611. doi:10.1016/j.asoc.2022.109611

work page doi:10.1016/j.asoc.2022.109611 2022
[25]

Y. Zhou, C. Huang, Q. Hu, J. Zhu, Y. Tang, Personalized learning full- path recommendation model based on LSTM neural networks, Inf. Sci. 444 (2018) 135–152

work page 2018
[26]

Graves, Long short-term memory, Supervised Seq

A. Graves, Long short-term memory, Supervised Seq. Labelling Recur- rent Neural Netw. (2012) 37–45

work page 2012
[27]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017)

work page 2017
[28]

Abdelrahman, Q

G. Abdelrahman, Q. Wang, B. Nunes, Knowledge tracing: A survey, ACM Computing Surveys 55 (11) (2023) 1–37. doi:10.1145/3569579

work page doi:10.1145/3569579 2023
[29]

Chang, H.-J

H.-S. Chang, H.-J. Hsu, K.-T. Chen, Modeling exercise relationships in E-learning: A unified approach, in: Proceedings of the 8th International Conference on Educational Data Mining (EDM), 2015, pp. 532–535

work page 2015
[30]

L. Yu, Y. Pian, Z. Shen, P. Chen, X. Li, SLP: A multi-dimensional and consecutive dataset from K-12 education, in: Proceedings of the 29thInternationalConferenceonComputersinEducation(ICCE),Asia- Pacific Society for Computers in Education, 2021, pp. 261–266

work page 2021
[31]

M. Feng, N. T. Heffernan, K. R. Koedinger, Addressing the assessment challenge with an online system that tutors as it assesses, User Modeling and User-Adapted Interaction 19 (3) (2009) 243–266. 19

work page 2009

[1] [1]

A. Sun, X. Chen, Online education and its effective practice: A research review, J. Inf. Technol. Educ. Res. 15 (2016) 157–190

work page 2016

[2] [2]

M. D. B. Castro, G. M. Tumibay, A literature review: efficacy of online learning courses for higher education institution using meta-analysis, Educ. Inf. Technol. 26 (2021) 1367–1385. 16

work page 2021

[3] [3]

A. D. Dumford, A. L. Miller, Online learning in higher education: ex- ploring advantages and disadvantages for engagement, J. Comput. High. Educ. 30 (2018) 452–465

work page 2018

[4] [4]

M. L. Bernacki, M. J. Greene, N. G. Lobczowski, A systematic review of research on personalized learning: Personalized by whom, to what, how, and for what purpose (s)?, Educ. Psychol. Rev. 33 (2021) 1675–1715

work page 2021

[5] [5]

O. O. Ayeni, N. M. Al Hamad, O. N. Chisom, B. Osawaru, O. E. Adewusi, AI in education: A review of personalized learning and ed- ucational technology, GSC Adv. Res. Rev. 18 (2024) 261–271

work page 2024

[6] [6]

A. H. Nabizadeh, J. P. Leal, H. N. Rafsanjani, R. R. Shah, Learning path personalization and recommendation methods: A survey of the state-of-the-art, Expert Syst. Appl. 159 (2020) 113596

work page 2020

[7] [7]

N. W. Rahayu, R. Ferdiana, S. S. Kusumawardani, A systematic review of learning path recommender systems, Educ. Inf. Technol. 28 (2023) 7437–7460

work page 2023

[8] [8]

D. Shi, T. Wang, H. Xing, H. Xu, A learning path recommen- dation model based on a multidimensional knowledge graph frame- work for e-learning, Knowledge-Based Systems 195 (2020) 105618. doi:10.1016/j.knosys.2020.105618

work page doi:10.1016/j.knosys.2020.105618 2020

[9] [9]

Huang, Q

Z. Huang, Q. Liu, C. Zhai, Y. Yin, E. Chen, W. Gao, G. Hu, Exploring multi-objective exercise recommendations in online education systems, in: In Proc. ACM CIKM Conf., 2019, pp. 1261–1270

work page 2019

[10] [10]

X. Chen, J. Shen, W. Xia, J. Jin, Y. Song, W. Zhang, W. Liu, M. Zhu, R. Tang, K. Dong, Set-to-sequence ranking-based concept-aware learn- ing path recommendation, in: In Proc. AAAI Conf., Vol. 37, 2023, pp. 5027–5035

work page 2023

[11] [11]

Q. Li, W. Xia, L. Yin, J. Shen, R. Rui, W. Zhang, X. Chen, R. Tang, Y. Yu, Graph enhanced hierarchical reinforcement learning for goal- oriented learning path recommendation, in: In Proc. ACM CIKM Conf., 2023, pp. 1318–1327. 17

work page 2023

[12] [12]

Zhang, S

H. Zhang, S. Shen, B. Xu, Z. Huang, J. Wu, J. Sha, S. Wang, Item- difficulty-aware learning path recommendation: From a real walking perspective, in: In Proc. ACM SIGKDD Conf., 2024, pp. 4167–4178

work page 2024

[13] [13]

G. Luo, H. Gu, X. Dong, D. Zhou, HA-LPR: A highly adaptive learning path recommendation, Educ. Inf. Technol. 30 (2025) 14597–14627

work page 2025

[14] [14]

X. Yu, S. Yang, Z. Wang, S. Song, H. Ma, Z. Cao, X. Zhang, LIGHT: Enhancinglearningpathrecommendationviaknowledgetopology-aware sequence optimization, in: In Proc. ACM SIGIR Conf., 2025, pp. 306– 315

work page 2025

[15] [15]

Cheng, Z

X. Cheng, Z. Zhang, J. Wang, L. Fang, C. He, Q. Guan, S. Pan, W. Luo, GraphRAG-induced dual knowledge structure graphs for personalized learning path recommendation, in: In Proc. AAAI Conf., Vol. 40, 2026, pp. 14610–14620

work page 2026

[16] [16]

Cheng, H

W. Cheng, H. Du, C. Li, E. Ni, L. Tan, T. Xu, Y. Ni, Uncertainty- aware knowledge tracing, in: In Proc. AAAI Conf., Vol. 39, 2025, pp. 27905–27913

work page 2025

[17] [17]

Order Matters: Sequence to sequence for sets

O. Vinyals, S. Bengio, M. Kudlur, Order matters: Sequence to sequence for sets, arXiv preprint arXiv:1511.06391 (2015)

work page Pith review arXiv 2015

[18] [18]

Z. Li, A. Sun, C. Li, DiffuRec: A diffusion model for sequential recom- mendation, ACM Trans. Inf. Syst. 42 (2023) 1–28

work page 2023

[19] [19]

J. Ho, A. Jain, P. Abbeel, Denoising diffusion probabilistic models, Ad- vances in neural information processing systems 33 (2020) 6840–6851

work page 2020

[20] [20]

Birjali, A

M. Birjali, A. Beni-Hssane, M. Erritali, A novel adaptive E-learning model based on Big Data by using competence-based knowledge and social learner activities, Applied Soft Computing 69 (2018) 14–32

work page 2018

[21] [21]

M. Zare, C. Pahl, H. Rahnama, M. Nilashi, A. Mardani, O. Ibrahim, H. Ahmadi, Multi-criteria decision making approach in E-learning: A systematic review and classification, Applied Soft Computing 45 (2016) 108–128

work page 2016

[22] [22]

Dwivedi, V

P. Dwivedi, V. Kant, K. K. Bharadwaj, Learning path recommendation based on modified variable length genetic algorithm, Educ. Inf. Technol. 23 (2018) 819–836. 18

work page 2018

[23] [23]

Q. Liu, S. Tong, C. Liu, H. Zhao, E. Chen, H. Ma, S. Wang, Exploiting cognitive structure for adaptive learning, in: In Proc. ACM SIGKDD Conf., 2019, pp. 627–635

work page 2019

[24] [24]

Gavrilovic, T

N. Gavrilovic, T. Sibalija, D. Domazet, Design and implementation of discrete Jaya and discrete PSO algorithms for automatic collaborative learning group composition in an e-learning system, Applied Soft Com- puting 129 (2022) 109611. doi:10.1016/j.asoc.2022.109611

work page doi:10.1016/j.asoc.2022.109611 2022

[25] [25]

Y. Zhou, C. Huang, Q. Hu, J. Zhu, Y. Tang, Personalized learning full- path recommendation model based on LSTM neural networks, Inf. Sci. 444 (2018) 135–152

work page 2018

[26] [26]

Graves, Long short-term memory, Supervised Seq

A. Graves, Long short-term memory, Supervised Seq. Labelling Recur- rent Neural Netw. (2012) 37–45

work page 2012

[27] [27]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017)

work page 2017

[28] [28]

Abdelrahman, Q

G. Abdelrahman, Q. Wang, B. Nunes, Knowledge tracing: A survey, ACM Computing Surveys 55 (11) (2023) 1–37. doi:10.1145/3569579

work page doi:10.1145/3569579 2023

[29] [29]

Chang, H.-J

H.-S. Chang, H.-J. Hsu, K.-T. Chen, Modeling exercise relationships in E-learning: A unified approach, in: Proceedings of the 8th International Conference on Educational Data Mining (EDM), 2015, pp. 532–535

work page 2015

[30] [30]

L. Yu, Y. Pian, Z. Shen, P. Chen, X. Li, SLP: A multi-dimensional and consecutive dataset from K-12 education, in: Proceedings of the 29thInternationalConferenceonComputersinEducation(ICCE),Asia- Pacific Society for Computers in Education, 2021, pp. 261–266

work page 2021

[31] [31]

M. Feng, N. T. Heffernan, K. R. Koedinger, Addressing the assessment challenge with an online system that tutors as it assesses, User Modeling and User-Adapted Interaction 19 (3) (2009) 243–266. 19

work page 2009