Pith. sign in

REVIEW 3 major objections 5 minor 45 references

Language Representation Favored Zero-Shot Cross-Domain Cognitive Diagnosis

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read A model that describes students, exercises and concepts in text can diagnose mastery in entirely new subjects and platforms without retraining, sometimes matching models trained on the target data.

desk verdict A genuinely new representation idea and solid engineering, but the zero-shot claim is not tested because target-domain labels are baked into the profiles at inference. read the letter →

arxiv 2501.13943 v1 pith:6U6PAKMY submitted 2025-01-18 cs.CL cs.AIcs.CYcs.LG

classification cs.CLcs.AIcs.CYcs.LG
keywords cognitivediagnosiszero-shotlearningcross-domaintransfertextualprofileslanguage-cognitivemappingtextembeddingsstudentscorepredictionintelligenteducationsystems
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

Cognitive diagnosis models normally learn ID embeddings that only work in the domain where they were trained, so a new subject or platform means a new model. This paper proposes LRCD, which replaces IDs with textual cognitive profiles: concept names for concepts, concept names plus average accuracy for exercises, and the student's interaction records for students. These profiles are embedded into a unified language space and projected into cognitive-diagnosis space by small learned mappers, then fed to an ordinary CDM. The paper's central claim is that a model trained this way on source domains performs well zero-shot in completely different target domains, in some experiments reaching competitive performance with classic CDMs trained on the full target response data. If true, one trained model could serve new subjects and platforms without retraining.

What carries the argument

The load-bearing mechanism is the pair of textual cognitive profiles and language-cognitive mappers. Profiles are constructed by three rules: a concept profile is just the concept name; an exercise profile appends the exercise's average correct rate (ACR) to the names of its related concepts; a student profile is the set of textual profiles of that student's interactions, each interaction being the exercise's concept names, its ACR, and the student's score. These texts are vectorized by a frozen text-embedding module, with each student represented by mean pooling over their interactions, and the mappers $\mathcal{F}_s,\mathcal{F}_e,\mathcal{F}_c$ (two linear layers with ReLU) project the language-space vectors to cognitive-space vectors. Training is supervised by a binary cross-entropy loss over source domains, and zero-shot inference applies the frozen mappers to target-domain profiles. The framework is model-agnostic and is demonstrated with OR-KaNCD, KaNCD, KSCD, and MIRT as the integrated CDM $\mathcal{M}_{\text{CD}}$.

What would settle it

Embed the target-domain concepts in a language not seen in training (e.g., translate the concept names) and rerun the same frozen mappers; if AUC collapses to near chance while the original target still scores high, the claimed domain-invariant mapping depends on lexical overlap rather than on a transferable cognitive representation.

Watch

Extended reading notes

Core claim

LRCD asserts that the barrier to cross-domain cognitive diagnosis is not the diagnosis task itself but the input representation. By writing each entity as text—a concept's name, an exercise's related concept names plus its average correct rate, a student's list of interaction records—and embedding that text with a pretrained text-embedding module, all domains land in one language space. The proposed language-cognitive mappers (simple MLPs) then project these vectors into a cognitive space where any existing CDM can consume them. Since the mapping must apply to unseen entities, the paper defines zero-shot inference for a target domain by freezing the trained mappers and running them on the target's textual profiles. Experiments on three real datasets report that LRCD reaches at least 97.30% of oracle AUC in subject-level transfer and at least 94.91% in platform-level transfer, and in several settings its AUC is competitive with NCDM trained on target data.

Load-bearing premise

The load-bearing premise is that the mapping from language embeddings to cognitive vectors learned on source domains remains valid in a completely new domain, and that target-domain response logs are available to build the textual profiles, even though the model is 'zero-shot' in not training on those logs.

Editorial extensions

If this is right

  • A single LRCD model can diagnose target subjects on the same platform with zero overlap, reaching at least 97.30% of oracle AUC in the reported subject-level experiments.
  • A model trained on one platform's math logs can diagnose math on another platform, reaching at least 94.91% of oracle AUC in the reported platform-level experiments.
  • Source domains from science subjects transfer better than humanities sources, and higher-education sources transfer better downward; LRCD's unified space exposes these regularities.
  • The architecture can absorb different text-embedding modules and different CDMs; the authors recommend OpenAI-3-large and OR-KaNCD, while BERT remains serviceable under resource constraints.
  • A student's diagnosis can be edited by adding hypothetical interactions to the profile, letting a student preview the effect of practicing a concept without retraining or changing the response data.

Reading between the lines

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

  • If the mapper is the only transferable component, a sharper test is to hold the text-embedding module constant and swap concept-name language between source and target; large degradation would show the transfer rides on semantic similarity of names rather than a domain-general cognitive signal.
  • The profile construction needs labels in the target domain to compute ACR and student interaction records, so a truly cold target with no prior response logs remains outside the method's stated zero-shot scope.
  • The same unified-space idea could be applied to interleaving diagnosis with knowledge tracing or to cross-course recommendation, since both tasks share the same student-exercise-concept structure; the paper does not pursue these applications.
  • Science-to-science and higher-to-lower transfer results suggest a hierarchy of learnability across domains; whether that hierarchy reflects data size, concept granularity, or inherent subject structure is left open.
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 / 5 minor

Summary. The paper proposes LRCD, a framework for what it calls zero-shot cross-domain cognitive diagnosis (ZSCD). LRCD converts students, exercises, and concepts into textual profiles derived from response logs, embeds these profiles with a text-embedding model, learns per-entity language-to-cognitive mappers on source domains, and then applies the frozen mappers to a target domain. The paper reports AUC and DOA values showing that LRCD approaches or matches classic CDMs trained on full target-domain data, and it claims that LRCD can transfer across subjects and platforms. The central technical idea—using text embeddings as a shared representation across domains—is plausible and the authors provide a code release and ten-seed experiments. However, the experimental protocol contradicts the paper's own zero-shot definition because target-domain training labels are embedded in the profiles used at inference, and the source-training student representation includes the label of the interaction being predicted. These issues are load-bearing for the paper's headline claim.

Significance. If the reported results were valid, the paper would make a significant practical contribution: a single model trained on source domains could diagnose students in arbitrary new subjects or platforms without target-domain training. The use of textual cognitive profiles and learned language-cognitive mappers is a reasonable way to seek a domain-invariant representation, and integrating the approach with existing CDMs is a useful practical design. The authors also share code and report multiple random seeds, which are good practices. However, the significance is undermined by two flaws: target-domain labels are used at inference despite the stated zero-shot setting, and the source-training objective can exploit the label token embedded in the student profile. The ablation study further reports implausibly high performance when all textual profiles are replaced by random vectors. As a result, the experiments do not establish that a source-trained mapper generalizes to unseen target domains, and the central claim is not supported.

major comments (3)
  1. [Section 3 vs. Section 4.3 / Section 5.1, Eqs. (2)-(4)] The inference procedure violates the paper's own definition of zero-shot cross-domain cognitive diagnosis. Section 3 defines ZSCD as having 'no overlapping students, exercises and concepts' and 'lacks any information about the target domain.' Section 4.3, however, states that at zero-shot inference the model 'tackle[s] the response logs in the train data of target domains into textual cognitive profiles.' Equation (2) puts ACR_ej, the average target-train correctness, into the exercise profile, and Eq. (3) puts y_ij, the student's actual score, into the student profile; Eq. (4) then mean-pools the student's target-train interactions. Section 5.1 confirms that 'we use the training data in the target domain to infer the students' mastery levels and evaluate the results on the test data.' Thus the reported AUCs in Tables 2 and 3 are obtained with target-domain labels available as input features; they are transductive or target-supervised results, not zero-shot results. This invalidates the abstract's claim that 'training LRCD on most real-world datasets can achieve commendable zero-shot performance.'
  2. [Section 4.1, Eqs. (3)-(4)] During source training, the student representation used to predict y_ij includes y_ij itself. Equation (3) defines each interaction profile I_ij as the text '[{c_k | Q_j,k = 1}, ACR_ej, y_ij]', and Eq. (4) sets the student's language-space vector h_si to the mean of the text embeddings of all of that student's interactions. When the model predicts y_ij for a training interaction, that same y_ij token is present in the mean-pooled student representation. Consequently, the student mapper can learn to detect and copy the label token rather than to compute a domain-invariant cognitive mapping. The paper provides no ablation or analysis that removes the label token from the student profile to show that the reported cross-domain performance is not driven by this shortcut.
  3. [Section 5.5, Table 4] The ablation LRCD-w/o-TCP is internally inconsistent with the claim that textual cognitive profiles are essential. This variant replaces every textual profile with a random vector sampled from a standard normal distribution, yet it still achieves AUC values of 75.74, 80.67, and 80.38 on PB-M, PB-C, and EHG-C, respectively, far above the 50% random baseline and close to the full LRCD. A model with random, information-free student, exercise, and concept representations should not be able to reach such performance unless there is additional label leakage in the training pipeline or the evaluation protocol does not reflect the diagnostic task. The authors should explain what information remains in this variant and why it performs so close to the full model; as reported, the ablation undermines the interpretation of the main results.
minor comments (5)
  1. [Section 5.5] LRCD-w/o-LCM is not a clean ablation of the mapper because it substitutes Bert embeddings for the default OpenAI-3-large to avoid memory limits, and the authors report that the variant still exceeds memory limits in some cases. The reported comparison therefore conflates the effect of the mapper with the effect of changing the text-embedding model.
  2. [Section 5.2] The transferability conclusion that science subjects transfer better than humanities subjects is confounded by differences in dataset size, average correct rate, and number of students across subjects. The paper itself acknowledges that 'we cannot rule out the influence of the data size,' so this finding should be presented as a speculative observation rather than a general insight.
  3. [Section 5.1, Implementation Details] The text says 'OR-KaNCD, proposed in ORCDF [36]' but reference [36] is the NeuralCD paper; ORCDF is reference [24]. Please correct this citation.
  4. [Section 4.3, Eq. (6)] The notation h_c in Eq. (6) is used without a prior definition; earlier, concept-level vectors are written as h_ck. Please clarify how a single concept representation is selected or aggregated when an exercise is associated with multiple concepts.
  5. [Section 4.4] The statement that the time complexity of the mappers is 'approximately O(d_l d)' is imprecise because the mapper is an MLP with two linear layers whose cost also depends on the intermediate hidden dimension. Please state the complexity in terms of the actual architecture used.

Circularity Check

2 steps flagged · score 6.0 of 10

LRCD's pipeline feeds the label being predicted (y_ij) into the student profile, and builds target-domain 'zero-shot' profiles from target training labels, so the reported transfer results are partly forced by construction.

  1. self definitional [Section 4.1, Students' Profiles (Eqs. 3-4); Section 4.3, Eq. 7]
    "I_{ij} = [{c_k | Q_{j,k} = 1}, ACR_{e_j}, y_{ij}], P_{s_i} = {I_{i0},..., I_{ij},...} ... h^{(l)}_{s_i} = ∑_j TEM(I_{ij})/|P_{s_i}| ... L_{R_m} = -∑_{(s,e,c,y_{se})∈R_m} [y_{se} log \hat{y}_{se} + (1-y_{se}) log(1-\hat{y}_{se})]"

    The same y_ij that Eq. 7 uses as the supervised target is a token inside I_ij, and Eq. 4 mean-pools I_ij into the student vector h_s_i that produces \hat{y}_ij. Therefore the model's input for a source training sample already contains that sample's label; the student mapper can learn to copy y_ij rather than a domain-invariant cognitive representation. The source-training objective is self-referential by construction, so success on source domains cannot establish the language-to-cognitive mapper generalization claimed for ZSCD.

  2. fitted input called prediction [Section 4.1 Eq. 2; Section 4.3 Zero-Shot Inference; Section 5.1 Evaluation]
    "ACR_{e_j} = 1/Z_j ∑_i y_{ij}, P_{e_j} = [{c_k | Q_{j,k}=1}, ACR_{e_j}] ... Then, we tackle the response logs in the train data of target domains into textual cognitive profiles ... During the inference phase, we use the training data in the target domain to infer the students' mastery levels and evaluate the results on the test data."

    The Introduction and Section 3 define ZSCD as 'lacks any information about the target domain' and having no overlapping students, exercises, or concepts. Yet target-domain inference constructs exercise profiles from ACR (the average of target-train y_ij) and student profiles from target-train responses that include y_ij. The reported zero-shot AUCs in Tables 2-3 therefore measure a transductive predictor given target labels as features, not a source-only model generalizing to unseen domains. The target ACR is a fitted summary of the target labels, and the test prediction is closely related to that summary by construction.

full rationale

The central derivation is not fully forced: the test labels themselves are not directly fed into the model for test predictions, and the language-cognitive mappers are fit only on source-domain data. However, the paper contains two concrete by-construction leakages. First, Eq. 3 places y_ij inside the student interaction profile, and Eq. 4 averages that profile into the student vector used to predict y_ij, so the source training objective is self-referential: the input already contains the output. Second, the supposed zero-shot target inference builds both student and exercise profiles from target training response logs, including target labels (Eqs. 2-3), despite the paper's definition of ZSCD as lacking any target information; Section 5.1 explicitly instructs using target training data during inference. Consequently, the headline 'zero-shot' results in Tables 2-3 do not test the claimed source-only generalization, and the high AUCs are partly attributable to target-label information embedded in the input profiles. The other candidate concerns (self-citations, post hoc subject-transferability insights, the LRCD-w/o-TCP ablation) are not load-bearing circularity: self-citations are not used to justify the core mapping, and the transferability observations are post hoc rather than part of the derivation. Overall this is substantial, construction-level leakage in the central claim, though not a full identity between output and input, hence a score of 6.

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

LRCD does not postulate new physical or conceptual entities such as a new force or conserved quantity. The 'textual cognitive profiles' and 'language-cognitive mappers' are model components, not independent invented entities with falsifiable handles outside the paper. The free parameters are hand-chosen hyperparameters and combination ratios that the central results depend on only weakly. The axioms are mostly domain assumptions about transferability and the validity of the evaluation surrogate; the two ad hoc axioms are the weakest because they concern the definition of zero-shot and the domain invariance of the mapper.

free parameters (4)
  • Data filtering thresholds = 5, 20, 50 responses
    Students with fewer than 5 (SLP), 20 (MOOC), 50 (EDM) responses are excluded; this hand choice affects profile quality and cross-domain comparability.
  • Multi-domain loss weight w_m = 1/|R_o|
    Loss weights across source domains are fixed to uniform weights by hand; the paper defers learned weights to future work.
  • Profile editing combination ratio = 7:3
    New profile and old profile are combined in a 7:3 ratio in the case study; no search or justification is given.
  • Mapper hidden dimensions = 512, 256
    MLP dimensions for the language-cognitive mappers are fixed to 512 and 256 across all methods; performance is not reported as a function of these.
assumptions (5)
  • domain assumption A frozen text-embedding module maps textual cognitive profiles from all domains into a shared language space where educational semantics are comparable.
    Used in Section 4.2 Eq. (4); profiles are embedded by OpenAI-3-large and treated as comparable across subjects and platforms.
  • ad hoc to paper The language-cognitive mapper trained on source domains generalizes to arbitrary target domains; the cognitive space is domain-invariant.
    Section 4.3 and zero-shot inference rely on this. Without it, the frozen mapper cannot place target students in the shared cognitive space.
  • ad hoc to paper Textual profiles constructed from target training labels are allowed at inference without violating the zero-shot problem definition.
    Section 5.1 uses ACR and student interaction scores from target training data to build profiles; the intro claims no target information is used.
  • domain assumption Mastery can be measured by score prediction accuracy on held-out response logs.
    Section 5.1 evaluation follows the standard surrogate: true mastery is unobservable, so AUC on response prediction is the metric.
  • standard math Standard supervised training and BCE loss on source logs is sufficient for learning mappers.
    Section 4.3 Eq. (7) uses binary cross-entropy; this is routine.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Language Representation Favored Zero-Shot Cross-Domain Cognitive Diagnosis." pith.science (2026). https://pith.science/paper/6U6PAKMY

@misc{pith2026250113943,
  author       = {Pith},
  title        = {Pith review of: Language Representation Favored Zero-Shot Cross-Domain Cognitive Diagnosis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6U6PAKMY}},
  note         = {Machine review of arXiv:2501.13943}
}
read the original abstract

Cognitive diagnosis aims to infer students' mastery levels based on their historical response logs. However, existing cognitive diagnosis models (CDMs), which rely on ID embeddings, often have to train specific models on specific domains. This limitation may hinder their directly practical application in various target domains, such as different subjects (e.g., Math, English and Physics) or different education platforms (e.g., ASSISTments, Junyi Academy and Khan Academy). To address this issue, this paper proposes the language representation favored zero-shot cross-domain cognitive diagnosis (LRCD). Specifically, LRCD first analyzes the behavior patterns of students, exercises and concepts in different domains, and then describes the profiles of students, exercises and concepts using textual descriptions. Via recent advanced text-embedding modules, these profiles can be transformed to vectors in the unified language space. Moreover, to address the discrepancy between the language space and the cognitive diagnosis space, we propose language-cognitive mappers in LRCD to learn the mapping from the former to the latter. Then, these profiles can be easily and efficiently integrated and trained with existing CDMs. Extensive experiments show that training LRCD on real-world datasets can achieve commendable zero-shot performance across different target domains, and in some cases, it can even achieve competitive performance with some classic CDMs trained on the full response data on target domains. Notably, we surprisingly find that LRCD can also provide interesting insights into the differences between various subjects (such as humanities and sciences) and sources (such as primary and secondary education).

Figures

Figures reproduced from arXiv: 2501.13943 by the authors.

Figure 1
Figure 1. An example of zero-shot cross-domain cognitive [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. An overview of our proposed LRCD framework. The left side (i.e., sub-figure (a)) provides an overview of the proposed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Overall student score prediction performance in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: t-SNE visualization of students’ mastery levels in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Case study of student profile editing. four renowned CDMs (i.e., MIRT [32], KSCD [23], KaNCD [36], OR-KaNCD [24]) for our experiments. As shown in [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: (a) Training time comparison with baselines. (b) [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Comparison of LRCD with different integrated CDM. [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Comparison with OR-KaNCD in the standard set [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Comparison with KaNCD in the standard setting. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: Comparison with KSCD in the standard setting. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 37 canonical work pages

  1. [1]

    Xiangzhi Chen, Le Wu, Fei Liu, Lei Chen, Kun Zhang, Richang Hong, and Meng Wang. 2023. Disentangling Cognitive Diagnosis with Limited Exercise Labels. In Advances in Neural Information Processing Systems 36 . New Orleans, LA

  2. [2]

    Yuxin Chen, Junfei Tan, An Zhang, Zhengyi Yang, Leheng Sheng, Enzhi Zhang, Xiang Wang, and Tat-Seng Chua. 2024. On Softmax Direct Preference Optimiza- tion for Recommendation. CoRR abs/2406.09215 (2024)

  3. [3]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies . Minneapolis, MN, 4171–4186

  4. [4]

    NHeffernan Ethan Prihar. 2023. EDM Cup 2023. https://kaggle.com/competitions/ edm-cup-2023

  5. [5]

    Weibo Gao, Qi Liu, Zhenya Huang, Yu Yin, Haoyang Bi, Mu-Chun Wang, Jianhui Ma, Shijin Wang, and Yu Su. 2021. RCD: Relation map driven cognitive diagnosis for intelligent education systems. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval . Virtual Event, 501–510

  6. [6]

    Weibo Gao, Qi Liu, Hao Wang, Linan Yue, Haoyang Bi, Yin Gu, Fangzhou Yao, Zheng Zhang, Xin Li, and Yuanjing He. 2024. Zero-1-to-3: Domain-Level Zero- Shot Cognitive Diagnosis via One Batch of Early-Bird Students towards Three Diagnostic Objectives. In Proceedings of the 38th AAAI Conference on Artificial Intelligence. Vancouver,Canada, 8417–8426

  7. [7]

    Weibo Gao, Hao Wang, Qi Liu, Fei Wang, Xin Lin, Linan Yue, Zheng Zhang, Rui Lv, and Shijin Wang. 2023. Leveraging Transferable Knowledge Concept Graph Embedding for Cold-Start Cognitive Diagnosis. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. Taipei, China, 983–992

  8. [8]

    Hengnian Gu, Zhiyi Duan, Pan Xie, and Dongdai Zhou. 2024. Modeling Balanced Explicit and Implicit Relations with Contrastive Learning for Knowledge Concept Recommendation in MOOCs. In Proceedings of the ACM on Web Conference 2024 . Singapore, Singapore, 3712–3722

Show all 45 references
  1. [9]

    H Tolga Kahraman, Seref Sagiroglu, and Ilhami Colak. 2010. Development of adaptive and intelligent web-based educational systems. In Proceedings of the 4th international conference on application of information and communication technologies. Halifax Nova Scotia, Canada, 1–5

  2. [10]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. 2015. Adam: A Method for Stochastic Optimiza- tion. In Proceedings of the 3rd International Conference on Learning Representations. San Diego, California

  3. [11]

    Jiatong Li, Qi Liu, Fei Wang, Jiayu Liu, Zhenya Huang, Fangzhou Yao, Linbo Zhu, and Yu Su. 2024. Towards the Identifiability and Explainability for Personalized Learner Modeling: An Inductive Paradigm. In Proceedings of the ACM on Web Conference 2024. Singapore, Singapore, 3420–3431

  4. [12]

    Jiatong Li, Fei Wang, Qi Liu, Mengxiao Zhu, Wei Huang, Zhenya Huang, Enhong Chen, Yu Su, and Shijin Wang. 2022. HierCDF: A Bayesian Network-based Hierarchical Cognitive Diagnosis Framework. In Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining...

  5. [13]

    Mingjia Li, Hong Qian, Jinglan Lv, Mengliang He, Wei Zhang, and Aimin Zhou

  6. [14]

    Sheng Li, Quanlong Guan, Liangda Fang, Fang Xiao, Zhenyu He, Yizhou He, and Weiqi Luo. 2022. Cognitive Diagnosis Focusing on Knowledge Concepts. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management. Atlanta, GA, 3272–3281

  7. [15]

    Qi Liu. 2021. Towards a New Generation of Cognitive Diagnosis. In Proceedings of 30th International Joint Conference on Artificial Intelligence . Montreal, Canada, 4961–4964

  8. [16]

    Qi Liu, Zhenya Huang, Yu Yin, Enhong Chen, Hui Xiong, Yu Su, and Guoping Hu. 2021. EKT: Exercise-Aware Knowledge Tracing for Student Performance Prediction. IEEE Transactions on Knowledge and Data Engineering 33, 1 (2021), 100–115

  9. [17]

    Shuo Liu, Junhao Shen, Hong Qian, and Aimin Zhou. 2024. Inductive Cognitive Diagnosis for Fast Student Learning in Web-Based Intelligent Education Systems. In Proceedings of the ACM on Web Conference 2024 . Singapore, Singapore, 4260– 4271

  10. [18]

    Shuo Liu, An Zhang, Guoqing Hu, Hong Qian, and Tat-seng Chua. 2024. Prefer- ence Diffusion for Recommendation. arXiv preprint arXiv:2410.13117 (2024)

  11. [19]

    Xiaohao Liu, Zhulin Tao, Jiahong Shao, Lifang Yang, and Xianglin Huang. 2022. Elimrec: Eliminating single-modal bias in multimedia recommendation. In Pro- ceedings of the 30th ACM International Conference on Multimedia. Lisboa, Portugal, 687–695

  12. [20]

    Yingjie Liu, Tiancheng Zhang, Xuecen Wang, Ge Yu, and Tao Li. 2023. New Development of Cognitive Diagnosis Models. Frontiers of Computer Science 17, 1 (2023), 171604

  13. [21]

    Zitao Liu, Qiongqiong Liu, Jiahao Chen, Shuyan Huang, and Weiqi Luo. 2023. simpleKT: A Simple But Tough-to-Beat Baseline for Knowledge Tracing. In Proceedings of the 11th International Conference on Learning Representations. Kigali, Rwanda

  14. [22]

    Yu Lu, Yang Pian, Ziding Shen, Penghe Chen, and Xiaoqing Li. 2021. SLP: A Multi-Dimensional and Consecutive Dataset from K-12 Education. In Proceedings of the 29th International Conference on Computers in Education . 261–266

  15. [23]

    Haiping Ma, Manwei Li, Le Wu, Haifeng Zhang, Yunbo Cao, Xingyi Zhang, and Xuemin Zhao. 2022. Knowledge-Sensed Cognitive Diagnosis for Intelligent Education Platforms. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management . Atlanta, GA, 1451–1460

  16. [24]

    Hong Qian, Shuo Liu, Mingjia Li, Bingdong Li, Zhi Liu, and Aimin Zhou. 2024. ORCDF: An Oversmoothing-Resistant Cognitive Diagnosis Framework for Stu- dent Learning in Online Education Systems. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Min...

  17. [25]

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. , 9 pages

  18. [26]

    Junhao Shen, Hong Qian, Wei Zhang, and Aimin Zhou. 2024. Symbolic Cognitive Diagnosis via Hybrid Optimization for Intelligent Education Systems. In Proceed- ings of the 38th AAAI Conference on Artificial Intelligence . Vancouver, Canada, 14928–14936

  19. [27]

    Xiaoteng Shen, Rui Zhang, Xiaoyan Zhao, Jieming Zhu, and Xi Xiao. 2024. PMG: Personalized Multimodal Generation with Large Language Models. InProceedings of the ACM on Web Conference 2024 . Singapore, Singapore, 3833–3843

  20. [28]

    Leheng Sheng, An Zhang, Yi Zhang, Yuxin Chen, Xiang Wang, and Tat-Seng Chua. 2024. Language Models Encode Collaborative Signals in Recommendation. arXiv preprint arXiv:2407.05441 (2024)

  21. [29]

    Jianwen Sun, Fenghua Yu, Sannyuya Liu, Yawei Luo, Ruxia Liang, and Xiaoxuan Shen. 2023. Adversarial Bootstrapped Question Representation Learning for Knowledge Tracing. In Proceedings of the 31st ACM International Conference on Multimedia. Ottawa, Canada, 8016–8025

  22. [30]

    Jianwen Sun, Fenghua Yu, Qian Wan, Qing Li, Sannyuya Liu, and Xiaoxuan Shen

  23. [31]

    Yang Sun, Fajie Yuan, Min Yang, Alexandros Karatzoglou, Li Shen, and Xiaoyan Zhao. 2022. Enhancing Top-N Item Recommendations by Peer Collaboration. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval . Madrid, Spa...

  24. [32]

    James B Sympson. 1978. A model for testing with multidimensional items. In Proceedings of the 1977 Computerized Adaptive Testing Conference . Minneapolis, MN

  25. [33]

    Zhulin Tao, Xiaohao Liu, Yewei Xia, Xiang Wang, Lifang Yang, Xianglin Huang, and Tat-Seng Chua. 2022. Self-supervised learning for multimedia recommenda- tion. IEEE Transactions on Multimedia (2022)

  26. [34]

    Laurens Van der Maaten and Geoffrey Hinton. 2008. Visualizing data using t-SNE. Journal of Machine Learning Research 9, 11 (2008)

  27. [35]

    Fei Wang, Qi Liu, Enhong Chen, Zhenya Huang, Yuying Chen, Yu Yin, Zai Huang, and Shijin Wang. 2020. Neural Cognitive Diagnosis for Intelligent Education Systems. In Proceedings of the 34th AAAI Conference on Artificial Intelligence . New York, NY

  28. [36]

    Fei Wang, Qi Liu, Enhong Chen, Zhenya Huang, Yu Yin, Shijin Wang, and Yu Su. 2023. NeuralCD: A General Framework for Cognitive Diagnosis. IEEE Transactions on Knowledge and Data Engineering 35, 8 (2023)

  29. [37]

    Wei Xu and Yuhan Zhou. 2020. Course video recommendation with multimodal information in online learning platforms: A deep learning framework. British Journal of Educational Technology 51, 5 (2020), 1734–1747

  30. [38]

    Shangshang Yang, Xiaoshan Yu, Ye Tian, Xueming Yan, Haiping Ma, and Xingyi Zhang. 2023. Evolutionary Neural Architecture Search for Transformer in Knowl- edge Tracing. In Advances in Neural Information Processing Systems 36 . New Orleans, LA

  31. [39]

    Jifan Yu, Mengying Lu, Qingyang Zhong, Zijun Yao, Shangqing= Tu, Zheng- shan Liao, Xiaoya Li, Manli Li, Lei Hou, Haitao Zheng, Juanzi Li, and Jie Tang

  32. [40]

    Yan Zhuang, Qi Liu, Zhenya Huang, Zhi Li, Shuanghong Shen, and Haiping Ma. 2022. Fully Adaptive Framework: Neural Computerized Adaptive Testing for Online Education. In Proceeddings of the 36th AAAI Conference on Artificial Intelligence. Virtual Event, 4734–4742

  33. [41]

    Yan Zhuang, Qi Liu, GuanHao Zhao, Zhenya Huang, Weizhe Huang, Zachary Pardos, Enhong Chen, Jinze Wu, and Xin Li. 2023. A Bounded Ability Estimation for Computerized Adaptive Testing. In Advances in Neural Information Processing Systems 36. New Orleans, LA

  34. [42]

    zero sample

    Jianhuan Zhuo, Jianxun Lian, Lanling Xu, Ming Gong, Linjun Shou, Daxin Jiang, Xing Xie, and Yinliang Yue. 2022. Tiger: Transferable Interest Graph Embedding for Domain-Level Zero-Shot Recommendation. In Proceedings of the 31st ACM International Conference on Information & Know...

  35. [2023]

    MoocRadar: A Fine-grained and Multi-aspect Knowledge Repository for Improving Cognitive Student Modeling in MOOCs. (2023)

  36. [2024]

    In Proceedings of the ACM on Web Conference 2024

    Interpretable Knowledge Tracing with Multiscale State Representation. In Proceedings of the ACM on Web Conference 2024 . 3265–3276

  37. [2025]

    Foundation Model Enhanced Derivative-Free Cognitive Diagnosis.Frontiers of Computer Science 19, 1 (2025), 191318

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.