Pith. sign in

REVIEW 3 major objections 5 minor 49 references

A Forced-Choice Neural Cognitive Diagnostic Model of Personality Testing

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Neural model tops forced-choice personality test ranking benchmarks.

desk verdict A plausible first neural CDM for forced-choice personality tests, with a real but fixable gap in the monotonicity argument and thin statistical evidence. read the letter →

arxiv 2507.15013 v1 pith:S3YBSEBS submitted 2025-07-20 cs.AI

Xiaoyu Li , Jin Wu , Shaoyang Guo , Haoran Shi , Chanjin Zheng This is my paper
classification cs.AI
keywords forced-choicetestscognitivediagnosispersonalityassessmentdeeplearningitemresponsetheorymonotonicityrankinglossinterpretability
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

This paper tries to show that a deep learning model, FCNCD, can score forced-choice personality tests—where people rank or pick among equally attractive statements—more accurately than existing statistical and neural baselines while also returning interpretable trait estimates. The authors argue that standard item response theory models for forced-choice data are slow, linear, and limited, and that existing neural cognitive diagnosis models cannot handle forced-choice item blocks. FCNCD embeds participants and items, applies nonlinear feature mapping, models the interaction between ability, difficulty, and discrimination, and optimizes a rank-difference-weighted Bayesian personalized ranking loss. On two real datasets (MAP, BFI) and one simulated dataset (sim-mole), the paper reports that FCNCD achieves the best overall pairwise and listwise rank accuracy, and that its estimated trait scores agree with ranked responses on an interpretability measure. The practical stake is that personnel selection and career development could use one automated, interpretable model for the common forced-choice formats without faking-prone Likert scales.

What carries the argument

The machinery has five parts. (1) Embeddings: each participant is a one-hot vector times a trainable tensor, giving a proficiency vector per dimension; each item has separate difficulty and discrimination embedding vectors. (2) Nonlinear mapping: three sigmoid layers transform proficiency, difficulty, and discrimination into higher-dimensional features. (3) Interaction function: the core identity is $x = \mathbf{h}^{\mathrm{disc}}_2 \odot (\mathbf{h}^{\mathrm{prof}}_2 - \mathbf{h}^{\mathrm{diff}}_2)$, which multiplies a discrimination feature by the gap between participant proficiency and item difficulty, mirroring the 2PL item response function. (4) Monotonicity layer: two fully connected layers with nonnegative weight matrices $W_4$ and $W_5$ are intended to make higher latent trait produce higher predicted scores. (5) Rank-difference-weighted BPR loss: each within-block pair contributes $-\ln\sigma(\lambda (r_i - r_j)(y_i - y_j))$, weighting pairs by their rank difference, which adapts the recommender-system BPR loss to forced-choice ranking. A DOA metric quantifies interpretability by checking whether dimensions with higher estimated ability also have higher summed ranked scores.

What would settle it

For a trained FCNCD model, fix the item embeddings of any single item, vary one participant's proficiency value on that item's dimension across its observed range while holding other dimensions fixed, and plot the predicted score; if the curve ever decreases as proficiency increases, the monotonicity guarantee is violated. The paper does not report such a per-item monotonicity check, and the same test applied to FCNCD_MO, the unconstrained variant, would quantify how much non-negativity actually enforces.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a neural cognitive diagnosis model can be adapted to forced-choice item blocks—PICK, RANK, and MOLE—by combining interpretable participant and item embeddings with a monotonicity-constrained interaction network and a pairwise ranking loss. The model claims to reconstruct each participant's latent traits on the measured dimensions and, from those traits, to predict the within-block ranking of items. In the experiments, FCNCD reports the highest PRA and LRA on the MAP and sim-mole datasets and a top LRA on BFI, with the suboptimal BFI PRA still competitive. The ablation study attributes the gain mostly to the nonlinear mapping layer and the modified BPR loss, while the nonnegative-weight monotonicity constraint contributes little to predictive accuracy but appears essential for interpretability: the unconstrained variant scores near random on the DOA consistency measure. If these results hold, FCNCD provides a deep-learning route to scoring all three common forced-choice formats while returning dimension-level trait estimates.

Load-bearing premise

The paper assumes that constraining the final two weight matrices to be nonnegative guarantees that a participant’s higher latent trait on a dimension always raises the predicted score of items measuring that dimension, but because the trait first passes through an unconstrained nonlinear layer, the net effect of raising the trait can still be a lower predicted score.

Editorial extensions

If this is right

  • One model can score PICK, RANK, and MOLE forced-choice formats without reformatting the response data, simplifying test scoring pipelines.
  • The rank-difference-weighted BPR loss improves listwise rank accuracy on real data compared with plain BPR and listwise losses, pointing to a better objective for forced-choice ranking.
  • Interpretable dimension-level trait estimates, validated by DOA, would let organizations use forced-choice test results not just as a ranking but as a diagnostic profile for selection and development.
  • Because neural baselines beat IRT baselines on RANK-3 blocks while IRT models remain competitive on the simulated MOLE-4 block, block format should influence model choice.

Reading between the lines

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

  • A direct monotonicity check on trained models would test whether the claimed interpretability guarantee holds; if it fails, the diagnostic trait values may still be useful but the stated mechanism would need revision.
  • The architecture could transfer to other ipsative assessments—values, attitudes, situational judgment—where a Q-matrix linking items to dimensions is available.
  • The finding that the unconstrained variant matches FCNCD in accuracy but not interpretability suggests the nonnegativity constraint acts as a regularizer for interpretability rather than a performance booster, which could be tested by replacing it with soft penalties.
  • Adding response times and item text as embedding inputs is a natural extension; the expected gains are testable but not yet measured.
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 FCNCD, a deep-learning model for cognitive diagnosis of forced-choice personality tests. Participants and items are represented by embeddings; item difficulty and discrimination are embedded, all three are passed through nonlinear mapping layers, and an interaction function combines proficiency, difficulty, and discrimination before two fully connected layers with nonnegative weights. The model is trained with a weighted BPR loss that adapts pairwise ranking to RANK and MOLE block formats. The authors evaluate FCNCD on two real datasets (MAP, BFI) and one simulated dataset (sim-mole), report pairwise and listwise rank accuracies, ablation studies, and a 'degree of agreement' (DOA) analysis intended to support interpretability of the estimated trait values.

Significance. If the claims are substantiated, FCNCD would be a useful unification of forced-choice ranking prediction and cognitive diagnostic trait estimation, addressing a real gap: most existing deep cognitive diagnosis models are designed for binary or Likert responses rather than forced-choice blocks. The inclusion of a simulated dataset with known latent trait parameters is a strength, as is the systematic ablation of the nonlinear mapping, loss function, and monotonicity constraint. The ranking results in Table 2 are competitive, and the DOA analysis is a sensible first step toward interpretability. However, the central interpretability claim is not currently supported by a rigorous monotonicity guarantee, and the DOA evaluation does not independently validate the trait estimates. These issues are fixable within the scope of the manuscript, so the paper warrants revision rather than rejection.

major comments (3)
  1. [4.3, Eqs. (5), (8)–(10)] The claim that constraining W4 and W5 to be nonnegative enforces the monotonicity assumption is not proven. The final score is y = φ(W5 φ(W4 x + b4) + b5) with x = h_disc_2 ⊙ (h_prof_2 − h_diff_2). Nonnegativity of W4 and W5 makes y nondecreasing in the intermediate x, but h_prof_2 = φ(W1 h_prof_1 + b1) with W1 unconstrained (Eq. 5), so the derivative of y with respect to the latent trait h_prof_1 includes the term diag(φ′(W1 h_prof_1 + b1)) W1, which can have negative entries. Moreover, h_prof_1 is selected from the participant embedding W_s in Eq. (1), which is also unconstrained. Thus the architecture does not guarantee that a higher estimated ability yields a higher predicted score for an item measuring that dimension, and the interpretability interpretation in Section 5.8 is not supported by the monotonicity argument. The FCNCD_MO ablation shows only that the constraint affects DOA empirically; it does not provide the required guarantee. Please either prove monotonicity under the actual parameterization, constrain the earlier layers so that the whole chain is monotone, or substantially weaken the interpretability claim.
  2. [5.8, Eq. (26)] The DOA measure compares the model's estimated trait values with the observed rank sums used in training. Because the model is trained to predict exactly these rankings, a high DOA can partly reflect fit to the training labels rather than an independent confirmation that the trait estimates are valid psychological constructs. The paper does not state whether DOA is computed on held-out item blocks, and it does not use the sim-mole dataset's known true trait parameters to check whether the latent traits are recovered. Please report DOA separately for held-out blocks and, ideally, add a trait-recovery analysis on sim-mole (e.g., correlations between estimated and true trait values) to support the interpretability conclusion.
  3. [5.6, Table 2] The claim that FCNCD outperforms the benchmarks is not backed by any measure of uncertainty. Several margins are very small (e.g., MAP PRA: 0.598 vs. 0.596 for NCDM-R; BFI PRA: 0.679 vs. 0.681 for KaNCD-R, where FCNCD is actually worse). Since each experiment is repeated 10 times, the authors should report standard deviations or confidence intervals and, where appropriate, a significance test; otherwise the superiority claim in Section 5.6 is not supported by the reported point estimates.
minor comments (5)
  1. [4.3] There is a typo in the sentence 'we we constrain each element of W4 and W5'; please remove the duplicated word.
  2. [5.7] The text says FCNCD_MO 'will be included in future interpretability analyses,' but Figure 6 already includes FCNCD_MO; the wording should be made consistent with the actual experiments.
  3. [5.7, Table 3] In the sim-mole LRA column, FCNCD_MO (0.359) is slightly higher than FCNCD (0.357), so the sentence 'The complete FCNCD model outperforms all variants' is not strictly accurate for all metrics.
  4. [3] The sentence describing RANK blocks is ungrammatical: 'The RANK item block type ranks all items in item block zl are ranked by participants in full order.' Please rewrite.
  5. [1 and 4] The abstract and model description claim applicability to PICK, RANK, and MOLE, but the experiments only cover RANK and MOLE; a sentence explaining that PICK is a special case of the same framework, or an additional experiment, would substantiate the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central prediction is evaluated on held-out forced-choice rankings, the interpretability check is a data-based consistency analysis, and the self-citation is background only.

full rationale

The paper's main claim is that FCNCD predicts held-out item rankings in forced-choice blocks. This is supported by PRA/LRA on test blocks not used in training, so the outcome is not constructed from a fitted parameter. The interpretability evaluation (DOA, Eq. 26) compares estimated latent trait values with observed sums of ranked scores; while both quantities are derived from the same response logs, this is an external consistency check and is not a definitional reduction. The only self-citation, Ref. [25], is used as related work for deep CDMs on Likert data and does not support any load-bearing premise. The monotonicity guarantee claimed in Section 4.3 is questionable because W1 in Eq. (5) is unconstrained, so nonnegative W4/W5 do not ensure that higher latent traits yield higher predicted scores; however, that is a soundness/correctness limitation, not a circularity in the derivation chain. No fitted input is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known result is repackaged by definition. Hence the circularity score is 0.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The model relies on standard domain assumptions of cognitive diagnosis (valid Q-matrix, monotone item response functions, unidimensional items within blocks), plus hand-tuned hyperparameters. No new physical or conceptual entities are introduced.

free parameters (4)
  • lambda (BPR weighting coefficient) = 8 (MAP), 5 (BFI), 10 (sim-mole)
    Tuned per dataset in Section 5.5.1; selection appears to use the test split since no validation set is reported.
  • Embedding dimension d = 64
    Model hyperparameter set in Section 5.3; not derived from theory.
  • Hidden layer sizes = 256 (nonlinear mapping), 128 (fully connected)
    Architecture sizes specified in Section 5.3.
  • Batch size and learning rate = 256 and 1e-2 (MAP), 64 and 5e-3 (BFI), 32 and 5e-4 (sim-mole)
    Chosen per dataset in Sections 5.5.2-5.5.3.
assumptions (3)
  • domain assumption The Q-matrix correctly maps each item to the trait dimensions it measures.
    The model takes Q as given (Section 3); wrong Q would corrupt training and interpretability.
  • domain assumption Monotonicity: higher latent trait on a dimension increases the likelihood of ranking items from that dimension higher.
    Stated in Section 5.8 as the basis for interpretability; the nonnegativity constraints on W4 and W5 are meant to enforce it but do not cover the full path.
  • domain assumption Participant responses are generated by a latent-trait process where proficiency, difficulty, and discrimination interact as in the model equations.
    Equations (1)-(10) encode this interaction; for the simulated dataset, responses are generated by a Luce decision rule, which is a specific form of this assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Forced-Choice Neural Cognitive Diagnostic Model of Personality Testing." pith.science (2026). https://pith.science/paper/S3YBSEBS

@misc{pith2026250715013,
  author       = {Pith},
  title        = {Pith review of: A Forced-Choice Neural Cognitive Diagnostic Model of Personality Testing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/S3YBSEBS}},
  note         = {Machine review of arXiv:2507.15013}
}
read the original abstract

In the smart era, psychometric tests are becoming increasingly important for personnel selection, career development, and mental health assessment. Forced-choice tests are common in personality assessments because they require participants to select from closely related options, lowering the risk of response distortion. This study presents a deep learning-based Forced-Choice Neural Cognitive Diagnostic Model (FCNCD) that overcomes the limitations of traditional models and is applicable to the three most common item block types found in forced-choice tests. To account for the unidimensionality of items in forced-choice tests, we create interpretable participant and item parameters. We model the interactions between participant and item features using multilayer neural networks after mining them using nonlinear mapping. In addition, we use the monotonicity assumption to improve the interpretability of the diagnostic results. The FCNCD's effectiveness is validated by experiments on real-world and simulated datasets that show its accuracy, interpretability, and robustness.

Figures

Figures reproduced from arXiv: 2507.15013 by the authors.

Figure 1
Figure 1. A toy example of forced-choice cognitive diagnosis. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The framework of the proposed FCNCD model. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The impact of λ. 5.5.2 The effect of batch size The analysis of batch size is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The impact of batch size [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: The impact of train set ratio. divided into training and test sets in an 8:2 ratio by item block. An early-stopping mechanism was implemented: training stops and the best model is saved if evaluation metrics do not improve for 5 consecutive iterations. 5.6 Performance …
Figure 6
Figure 6. Figure 6: DOA results of models on three datasets. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Diagnosed results of two participants on MAP. [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 41 canonical work pages

  1. [1]

    Personnel selection.Journal of occupational and Organizational psychology, 74(4):441–472, 2001

    Ivan T Robertson and Mike Smith. Personnel selection.Journal of occupational and Organizational psychology, 74(4):441–472, 2001

    work page 2001

  2. [2]

    Psychometric sifting to efficiently select the right service employees.Managing Service Quality, 24(5):418–433, 2014

    John EG Bateson, Jochen Wirtz, Eugene Burke, and Carly Vaughan. Psychometric sifting to efficiently select the right service employees.Managing Service Quality, 24(5):418–433, 2014

    work page 2014

  3. [3]

    The career development inventory in review: Psychometric and research findings.Journal of Career Assessment, 4(2):171–188, 1996

    Mark L Savickas and Paul J Hartung. The career development inventory in review: Psychometric and research findings.Journal of Career Assessment, 4(2):171–188, 1996

    work page 1996

  4. [4]

    Psycho- metric evaluation of trauma and posttraumatic stress disorder assessments in persons with severe mental illness

    Kim T Mueser, Stanley D Rosenberg, Lindy Fox, Michelle P Salyers, Julian D Ford, and Patricia Carty. Psycho- metric evaluation of trauma and posttraumatic stress disorder assessments in persons with severe mental illness. Psychological assessment, 13(1):110, 2001

    work page 2001

  5. [5]

    Deep knowledge tracing.Advances in neural information processing systems, 28, 2015

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

    2015

  6. [6]

    Knowledge tracing: A survey.ACM Computing Surveys, 55(11):1–37, 2023

    Ghodai Abdelrahman, Qing Wang, and Bernardo Nunes. Knowledge tracing: A survey.ACM Computing Surveys, 55(11):1–37, 2023

    work page 2023

  7. [7]

    Neuralcd: a general framework for cognitive diagnosis.IEEE Transactions on Knowledge and Data Engineering, 2022

    Fei Wang, Qi Liu, Enhong Chen, Zhenya Huang, Yu Yin, Shijin Wang, and Yu Su. Neuralcd: a general framework for cognitive diagnosis.IEEE Transactions on Knowledge and Data Engineering, 2022

    work page 2022

  8. [8]

    Fuzzy Cognitive Diagnosis for Modelling Examinee Performance.ACM Transactions on Intelligent Systems and Technology, 9:1–26, 2018

    Qi Liu, Runze Wu, Enhong Chen, Guandong Xu, Yu Su, Zhigang Chen, and Guoping Hu. Fuzzy Cognitive Diagnosis for Modelling Examinee Performance.ACM Transactions on Intelligent Systems and Technology, 9:1–26, 2018

    work page 2018

Show all 49 references
  1. [9]

    A systematic literature review of personalized learning terms

    Atikah Shemshack and Jonathan Michael Spector. A systematic literature review of personalized learning terms. Smart Learning Environments, 7(1):33, 2020

    2020

  2. [10]

    Opq32r technical manual

    Anna Brown and Dave Bartram. Opq32r technical manual. 2011

    2011

  3. [11]

    Personality and job performance: the big five revisited.Journal of applied psychology, 85(6):869, 2000

    Gregory M Hurtz and John J Donovan. Personality and job performance: the big five revisited.Journal of applied psychology, 85(6):869, 2000

    2000

  4. [12]

    Predicting sales performance criteria with personality measures: The use of the general factor of personality, the big five and narrow traits.Human Performance, 26(2):126–149, 2013

    Thomas Sitser, Dimitri Van der Linden, and Marise Ph Born. Predicting sales performance criteria with personality measures: The use of the general factor of personality, the big five and narrow traits.Human Performance, 26(2):126–149, 2013

    2013

  5. [13]

    Minnesota multiphasic personality inventory; manual, revised

    Starke Rosecrans Hathaway and John Charnley McKinley. Minnesota multiphasic personality inventory; manual, revised. 1951

    1951

  6. [14]

    Validation and intensification of the sixteen personality factor questionnaire.Journal of Clinical Psychology, 12(3), 1956

    Raymond B Cattell. Validation and intensification of the sixteen personality factor questionnaire.Journal of Clinical Psychology, 12(3), 1956

    1956

  7. [15]

    The reliability and validity of normative and ipsative approaches in the measurement of personality.Journal of Occupational Psychology, 64(3):219–238, 1991

    Peter Saville and Eric Willson. The reliability and validity of normative and ipsative approaches in the measurement of personality.Journal of Occupational Psychology, 64(3):219–238, 1991

    1991

  8. [16]

    The impact of faking on employment tests: Does forced choice offer a solution?Human Performance, 13(4):371–388, 2000

    Douglas N Jackson, Victor R Wroblewski, and Michael C Ashton. The impact of faking on employment tests: Does forced choice offer a solution?Human Performance, 13(4):371–388, 2000

    2000

  9. [17]

    Eunike Wetzel, Susanne Frick, and Anna Brown. Does multidimensional forced-choice prevent faking? comparing the susceptibility of the multidimensional forced-choice format and the rating scale format to faking.Psychological Assessment, 33(2):156, 2021

    2021

  10. [18]

    Does forcing reduce faking? a meta-analytic review of forced-choice personality measures in high-stakes situations.Journal of Applied Psychology, 104(11):1347, 2019

    Mengyang Cao and Fritz Drasgow. Does forcing reduce faking? a meta-analytic review of forced-choice personality measures in high-stakes situations.Journal of Applied Psychology, 104(11):1347, 2019

    2019

  11. [19]

    Item response modeling of forced-choice questionnaires.Educational and Psychological Measurement, 71(3):460–502, 2011

    Anna Brown and Alberto Maydeu-Olivares. Item response modeling of forced-choice questionnaires.Educational and Psychological Measurement, 71(3):460–502, 2011

    2011

  12. [20]

    An irt approach to constructing and scoring pairwise preference items involving stimuli on different dimensions: The multi-unidimensional pairwise-preference model

    Stephen Stark, Oleksandr S Chernyshenko, and Fritz Drasgow. An irt approach to constructing and scoring pairwise preference items involving stimuli on different dimensions: The multi-unidimensional pairwise-preference model. Applied Psychological Measurement, 29(3):184–203, 2005

    2005

  13. [21]

    Neural Cognitive Diagnosis for Intelligent Education Systems.Proceedings of the AAAI Conference on Artificial Intelligence, 34:6153–6161, 2020

    Fei Wang, Qi Liu, Enhong Chen, Zhenya Huang, Yuying Chen, Yu Yin, Zai Huang, and Shijin Wang. Neural Cognitive Diagnosis for Intelligent Education Systems.Proceedings of the AAAI Conference on Artificial Intelligence, 34:6153–6161, 2020

    2020

  14. [22]

    ICD: A new interpretable cognitive diagnosis model for intelligent tutor systems.Expert Systems with Applications, 215:119309, 2023

    Tianlong Qi, Meirui Ren, Longjiang Guo, Xiaokun Li, Jin Li, and Lichen Zhang. ICD: A new interpretable cognitive diagnosis model for intelligent tutor systems.Expert Systems with Applications, 215:119309, 2023. 13 A Forced-Choice Neural Cognitive Diagnostic Model of Personalit...

    2023

  15. [23]

    Inductive cognitive diagnosis for fast student learning in web-based intelligent education systems

    Shuo Liu, Junhao Shen, Hong Qian, and Aimin Zhou. Inductive cognitive diagnosis for fast student learning in web-based intelligent education systems. InProceedings of the ACM on Web Conference 2024, pages 4260–4271, 2024

    2024

  16. [24]

    Deep cognitive diagnosis model for predicting students’ performance.Future Generation Computer Systems, 126:252–262, 2022

    Lina Gao, Zhongying Zhao, Chao Li, Jianli Zhao, and Qingtian Zeng. Deep cognitive diagnosis model for predicting students’ performance.Future Generation Computer Systems, 126:252–262, 2022

    2022

  17. [25]

    An interpretable polytomous cognitive diagnosis framework for predicting examinee performance.Information Processing & Management, 62(1):103913, 2025

    Xiaoyu Li, Shaoyang Guo, Jin Wu, and Chanjin Zheng. An interpretable polytomous cognitive diagnosis framework for predicting examinee performance.Information Processing & Management, 62(1):103913, 2025

    2025

  18. [26]

    Comparing simple scoring with irt scoring of personality measures: The navy computer adaptive personality scales.Applied Psychological Measurement, 39(2):144–154, 2015

    Frederick L Oswald, Amy Shaw, and William L Farmer. Comparing simple scoring with irt scoring of personality measures: The navy computer adaptive personality scales.Applied Psychological Measurement, 39(2):144–154, 2015

    2015

  19. [27]

    Comparing traditional and irt scoring of forced-choice tests.Applied Psychological Measurement, 39(8):598–612, 2015

    Pedro M Hontangas, Jimmy De La Torre, Vicente Ponsoda, Iwin Leenen, Daniel Morillo, and Francisco J Abad. Comparing traditional and irt scoring of forced-choice tests.Applied Psychological Measurement, 39(8):598–612, 2015

    2015

  20. [28]

    Development of information functions and indices for the ggum-rank multidimensional forced choice irt model.Journal of Educational Measurement, 55(3):357–372, 2018

    Seang-Hwane Joo, Philseok Lee, and Stephen Stark. Development of information functions and indices for the ggum-rank multidimensional forced choice irt model.Journal of Educational Measurement, 55(3):357–372, 2018

    2018

  21. [29]

    A 2plm-rank multidimensional forced-choice model and its fast estimation algorithm.Behavior Research Methods, pages 1–26, 2024

    Chanjin Zheng, Juan Liu, Yaling Li, Peiyi Xu, Bo Zhang, Ran Wei, Wenqing Zhang, Boyang Liu, and Jing Huang. A 2plm-rank multidimensional forced-choice model and its fast estimation algorithm.Behavior Research Methods, pages 1–26, 2024

    2024

  22. [30]

    Strengths and limitations of ipsative measurement.Journal of Occupational and Organizational Psychology, 69(1):49–56, 1996

    Helen Baron. Strengths and limitations of ipsative measurement.Journal of Occupational and Organizational Psychology, 69(1):49–56, 1996

    1996

  23. [31]

    A general item response theory model for unfolding unidimensional polytomous responses.Applied Psychological Measurement, 24(1):3–32, 2000

    James S Roberts, John R Donoghue, and James E Laughlin. A general item response theory model for unfolding unidimensional polytomous responses.Applied Psychological Measurement, 24(1):3–32, 2000

    2000

  24. [32]

    Item response theory models for ipsative tests with multidimensional pairwise comparison items.Applied Psychological Measurement, 41(8):600– 613, 2017

    Wen-Chung Wang, Xue-Lan Qiu, Chia-Wen Chen, Sage Ro, and Kuan-Yu Jin. Item response theory models for ipsative tests with multidimensional pairwise comparison items.Applied Psychological Measurement, 41(8):600– 613, 2017

    2017

  25. [33]

    A bayesian random block item response theory model for forced-choice formats.Educational and Psychological Measurement, 80(3):578–603, 2020

    HyeSun Lee and Weldon Z Smith. A bayesian random block item response theory model for forced-choice formats.Educational and Psychological Measurement, 80(3):578–603, 2020

    2020

  26. [34]

    Bpr: Bayesian personalized ranking from implicit feedback.arXiv preprint arXiv:1205.2618, 2012

    Steffen Rendle, Christoph Freudenthaler, Zeno Gantner, and Lars Schmidt-Thieme. Bpr: Bayesian personalized ranking from implicit feedback.arXiv preprint arXiv:1205.2618, 2012

    2012 arXiv

  27. [35]

    Learning to rank using gradient descent

    Chris Burges, Tal Shaked, Erin Renshaw, Ari Lazier, Matt Deeds, Nicole Hamilton, and Greg Hullender. Learning to rank using gradient descent. InProceedings of the 22nd international conference on Machine learning, pages 89–96, 2005

    2005

  28. [36]

    Learning to rank with nonsmooth cost functions.Advances in neural information processing systems, 19, 2006

    Christopher Burges, Robert Ragno, and Quoc Le. Learning to rank with nonsmooth cost functions.Advances in neural information processing systems, 19, 2006

    2006

  29. [37]

    Exploiting Non-Interactive Exercises in Cognitive Diagnosis

    Fangzhou Yao, Qi Liu, Min Hou, Shiwei Tong, Zhenya Huang, Enhong Chen, Jing Sha, and Shijin Wang. Exploiting Non-Interactive Exercises in Cognitive Diagnosis. InProceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, pages 2397–2405. Interna...

    2023

  30. [38]

    Enhancing cognitive diagnosis using un-interacted exercises: A collaboration-aware mixed sampling approach

    Haiping Ma, Changqian Wang, Hengshu Zhu, Shangshang Yang, Xiaoming Zhang, and Xingyi Zhang. Enhancing cognitive diagnosis using un-interacted exercises: A collaboration-aware mixed sampling approach. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, p...

    2024

  31. [39]

    Learning to rank: from pairwise approach to listwise approach

    Zhe Cao, Tao Qin, Tie-Yan Liu, Ming-Feng Tsai, and Hang Li. Learning to rank: from pairwise approach to listwise approach. InProceedings of the 24th international conference on Machine learning, pages 129–136, 2007

    2007

  32. [40]

    Listwise approach to learning to rank: theory and algorithm

    Fen Xia, Tie-Yan Liu, Jue Wang, Wensheng Zhang, and Hang Li. Listwise approach to learning to rank: theory and algorithm. InProceedings of the 25th international conference on Machine learning, pages 1192–1199, 2008

    2008

  33. [41]

    Adarank: a boosting algorithm for information retrieval

    Jun Xu and Hang Li. Adarank: a boosting algorithm for information retrieval. InProceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval, pages 391–398, 2007

    2007

  34. [42]

    From ranknet to lambdarank to lambdamart: An overview.Learning, 11(23-581):81, 2010

    Christopher JC Burges. From ranknet to lambdarank to lambdamart: An overview.Learning, 11(23-581):81, 2010. 14 A Forced-Choice Neural Cognitive Diagnostic Model of Personality Testing

    2010

  35. [43]

    Christopher J Soto and Oliver P John. The next big five inventory (bfi-2): Developing and assessing a hierarchical model with 15 facets to enhance bandwidth, fidelity, and predictive power.Journal of personality and social psychology, 113(1):117, 2017

    2017

  36. [44]

    Daniel Morillo, Iwin Leenen, Francisco J Abad, Pedro Hontangas, Jimmy de la Torre, and Vicente Ponsoda. A dominance variant under the multi-unidimensional pairwise-preference framework: Model formulation and markov chain monte carlo estimation.Applied Psychological Measurement...

    2016

  37. [45]

    A bayesian random effects model for testlets.Psychometrika, 64:153–168, 1999

    Eric T Bradlow, Howard Wainer, and Xiaohui Wang. A bayesian random effects model for testlets.Psychometrika, 64:153–168, 1999

    1999

  38. [46]

    A cognitive diagnosis model with nonlinear dependence between students and exercises.IEEE Transactions on Computational Social Systems, pages 1–12, 2025

    Yuhong Zhang, Zhihao Lin, Chenyang Bu, Kui Yu, Xuegang Hu, and Xindong Wu. A cognitive diagnosis model with nonlinear dependence between students and exercises.IEEE Transactions on Computational Social Systems, pages 1–12, 2025

    2025

  39. [47]

    Understanding the difficulty of training deep feedforward neural networks

    Xavier Glorot and Yoshua Bengio. Understanding the difficulty of training deep feedforward neural networks. InProceedings of the thirteenth international conference on artificial intelligence and statistics, pages 249–256. JMLR Workshop and Conference Proceedings, 2010

    2010

  40. [48]

    Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    I Loshchilov. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    2017 arXiv

  41. [49]

    Testing the conditional independence and monotonicity assumptions of item response theory

    Paul R Rosenbaum. Testing the conditional independence and monotonicity assumptions of item response theory. Psychometrika, 49:425–435, 1984. 15

    1984

Pith tools

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