Scalable Early Childhood Reading Performance Prediction

Derek Kosty; Eshed Ohn-Bar; Hank Fien; Michael Stoolmiller; Ola Ozernov-Palchik; Zanming Huang; Zhongkai Shangguan

REVIEW 4 major objections 9 minor 74 references

Scalable Early Childhood Reading Performance Prediction

T0 review · 4 major / 9 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper demonstrates that a simple MLP pre-trained over randomly masked inputs — MaskMLP — predicts first-grade reading progress better than imputation-based baselines, with the largest gains among students receiving a reading…

desk verdict Useful new dataset, but the binary label is defined by the control group's end-of-year average gain, which isn't known at prediction time—so the benchmark results don't yet support the paper's deployment story. read the letter →

arxiv 2412.10401 v1 pith:KELXOXX2 submitted 2024-12-05 cs.LG

Zhongkai Shangguan , Zanming Huang , Eshed Ohn-Bar , Ola Ozernov-Palchik , Derek Kosty , Michael Stoolmiller , Hank Fien This is my paper · ORCID

classification cs.LG

keywords earlyreadingpredictionself-supervisedpre-trainingmaskedautoencodingmissingdatatabularlearningeducationalinterventionECRIdatasetfluencyassessment

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

The paper's goal is to show that future reading progress of first graders can be predicted from fall assessments and classroom variables, even when nearly a third of the data is missing, and that a simple self-supervised trick beats stronger-looking baselines. It introduces a public benchmark, the Enhanced Core Reading Instruction (ECRI) dataset, covering 6,916 students across 44 schools, with outcome labels defined by whether a student's year-end gain in word identification and word attack exceeded the control group's average gain. The central empirical claim is that MaskMLP — an MLP pre-trained by randomly masking observed inputs and enforcing a cosine-embedding similarity between original and masked inputs — outperforms logistic regression, gradient-boosted trees, TabNet, VIME, SCARF, and MLPs trained with imputation or missing-indicator features. The gains are larger for students in the intervention group, where MaskMLP reaches 0.7704 accuracy on the word identification task under school-disjoint splits. If right, this gives educators a simple recipe for early risk identification without waiting for a child to fail, and a public dataset to test such models.

What carries the argument

The central object is MaskMLP, a multi-layer perceptron pre-trained with a self-supervised objective that does no imputation: missing features are marked with a sentinel value of −1, a random 25% of observed features are additionally masked, and the network is trained to maximize cosine similarity between embeddings of the original and masked inputs. After this pre-training, a classification head is attached and the network is fine-tuned with binary cross-entropy. The mechanism that carries the argument is the embedding-space prediction task: instead of filling missing entries in the input space, the model learns relations among variables under partial observation, which the paper argues regularizes training and yields more separable student-profile clusters.

What would settle it

Recompute the same experiments after re-defining the positive class using the median control-group gain, or using a random half of the control group's mean, and check whether MaskMLP's accuracy and AUC advantages over the MLP-indicator and VIME baselines persist.

Watch

Extended reading notes

Core claim

The paper claims that a simple MLP pre-trained with a random-masking cosine-embedding objective — MaskMLP — predicts whether first-grade students make above-average reading progress better than logistic regression, XGBoost, LightGBM, TabNet, VIME, SCARF, and standard MLP variants with zero-, mean-, or indicator-based missing-value handling, and that the gains are largest among students who received the reading intervention. On the intervention subset under school-disjoint splits, MaskMLP reaches 0.7704 accuracy and 0.7633 AUC for word identification, compared with 0.7539 and 0.7394 for the best MLP baseline, and similar improvements hold for the word attack task. The paper further argues that the pre-training step creates more separable student-profile clusters in the embedding space and that a simple cosine loss works better than more complex contrastive or reconstruction objectives for this partially observed tabular setting.

Load-bearing premise

The binary label of 'made sufficient reading progress' is defined as improvement greater than the control group's average gain, so if that average is noisy or unrepresentative, every model comparison inherits unreliable labels.

Editorial extensions

If this is right

MaskMLP's consistent advantage over zero-filling, mean-filling, and indicator-based MLPs suggests that embedding-space masked pre-training is a practical default for tabular educational data with heavy missingness.
The intervention subgroup is easier to predict, with all models improving by roughly 10% in accuracy, so intervention settings provide a cleaner signal for identifying which students still fail to progress.
The public ECRI benchmark allows future work to compare models for early reading prediction on a common dataset, including subgroup analyses by gender, at-risk status, and school socioeconomic level.
Feature-removal analysis indicates that initial word-identification score is the single most influential input, with classroom and teacher measures playing smaller but nonzero roles.

Reading between the lines

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

Because MaskMLP never imputes values, the same recipe could transfer to other tabular domains with structured missingness, such as medical records, attendance, or administrative data, not just reading assessments.
The dataset's label depends on the control group's average gain, so benchmark difficulty may shift across cohorts; re-reporting results with a fixed external threshold would make comparisons more stable.
MaskMLP's embedding clusters, shown through t-SNE, could be used for unsupervised student profiling, identifying subgroups that share response patterns before any labels are applied.
A direct test would be to run the same masked pre-training on a math or writing outcomes dataset; a similar accuracy gain would indicate the method is a general early-academic-skills tool rather than a reading-specific one.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

The ECRI dataset is a genuine new resource, and MaskMLP is a simple, sensible idea. But the paper's central benchmark task has a problem that undercuts the main claim.

The binary outcome is defined as "improved more than the control group's average gain over the school year" (Sec. 3.1). That control average is an end-of-year quantity. The paper never says whether it's computed globally, per fold, or per school. If it's global and includes test-fold control students, then labels depend on test outcomes and the evaluation leaks. If it's computed from training data only, the model is internally well-defined, but the deployment story in the abstract doesn't work: a teacher making predictions at the start of the year doesn't know the current cohort's control average, so the predicted class isn't computable. Either way, the target is a property of the cohort, not of the student. That's more serious than threshold noise; it changes what the benchmark measures.

What the paper does well: the dataset fills a real gap (6,916 students, 44 schools, public release), and the authors are honest about missing data and evaluation limitations. The comparisons are broad, and the MaskMLP gains are consistent, especially on the intervention subgroup, with a paired t-test against MLP giving p=0.009. The bias and feature-importance analyses are useful.

The other soft spots are comparatively minor: no error bars across seeds (admitted in the checklist), complete-case-only evaluation despite the missing-data focus, and hyperparameters chosen without a nested validation split. All addressable.

My bottom line: this deserves a serious referee because the dataset is valuable and the method is plausible, but the label definition needs to be fixed or explicitly reframed before the numbers can be trusted as prospective predictions. I'd be wary of citing the main accuracy comparisons until that's resolved.

Recommendation: send to peer review with a request for the authors to clarify the threshold's computation and to re-run the benchmark with a label that is computable at prediction time (e.g., based on a fixed norm or training-fold control average).

Referee Report

4 major / 9 minor

Summary. The paper introduces the Enhanced Core Reading Instruction (ECRI) dataset, a longitudinal tabular dataset of 6,916 first-grade students from 44 schools, and proposes MaskMLP, a self-supervised MLP pre-training strategy that randomly masks observed features and aligns the original and masked embeddings via a cosine embedding loss. The authors evaluate MaskMLP against several baselines (logistic regression, XGBoost, LightGBM, TabNet, VIME, SCARF, and MLP variants with different missing-data strategies) on binary classification of word identification and word attack improvement. The binary label is defined as whether a student's gain exceeds the average gain of the control group. Experiments are reported under school-split and student-split protocols, with a separate analysis on the intervention subgroup. The paper claims that MaskMLP consistently achieves the best performance, especially on the intervention subset, and also provides feature importance, bias characterization, and embedding visualization analyses.

Significance. If the evaluation is valid, the ECRI dataset would be a valuable public resource for research on early reading prediction and educational ML, filling a clear gap in available datasets. The MaskMLP method is simple and potentially useful for tabular data with missing values, and the paper's comprehensive comparison with multiple baselines is a strength. The authors also release data and code, and they conduct a bias analysis, which is commendable. However, the significance is substantially undercut by the cohort-relative label definition, which may make the reported task ill-posed for the claimed early prediction use case, and by the exclusion of missing-data samples from fine-tuning and evaluation, which may bias the reported numbers. The statistical evidence for the central claim of consistent improvement is also weak, with borderline p-values and no across-seed error bars. These issues need to be resolved before the contributions can be fully assessed.

major comments (4)

[Sec. 5.1] The binary target is defined as whether a student's improvement exceeds 'the average performance improvement of the control group over a school year' (Sec. 3.1) and 'higher than the average in the control group' (Sec. 4.1). This makes the label a cohort-relative quantity. The paper does not state whether the threshold is computed globally, per school, per fold, or whether test-fold control outcomes are used. If the threshold uses test-fold outcomes, the labels are partially defined by the outcomes being predicted. If it is computed only from training data, the model still cannot be deployed for the 'early prediction' claimed in the abstract, because an educator at the start of the year does not know the control group's mean end-of-year gain. The authors should specify the exact computation and provide a sensitivity analysis using an absolute or a fixed reference threshold, or otherwise justify that the cohort-relative label is a well-defined prediction target for the claimed use case.
[Sec. 3.1, Sec. 5] The paper states that samples with missing data are retained for pre-training but excluded from fine-tuning, evaluation, and analysis. With 30.48% of entries missing, this exclusion removes a substantial and potentially non-random subset of students (e.g., those frequently absent). If missingness is informative, the reported accuracy and AUC are estimated on a biased complete-case sample and do not reflect the population in which the model would be deployed. The authors should evaluate MaskMLP and at least one strong baseline on the full test set using their missing-value handling (e.g., the -1 indicator approach) and compare these results with the complete-case results, along with a discussion of the missingness mechanism.
[Sec. 5.1, Checklist 3(c)] The statistical evidence for MaskMLP's advantage is weak. The paired t-test over 5 folds has very low power, and the reported p-values for comparisons against VIME and SCARF are 0.0517 and 0.0521, respectively, which are not significant at the conventional 0.05 level. On the full student set in the school-split word identification task, the accuracy difference between MaskMLP (0.6726) and the MLP (Indicator) baseline (0.6710) is only 0.0016. The checklist also confirms that no error bars across random seeds are reported. The authors should report confidence intervals (e.g., bootstrap over folds) and results from multiple random seeds, and clearly state which MLP variant is used in the t-test, so that the 'consistently outperforms' claim is supported by appropriate uncertainty quantification.
[Table 1, Sec. 5.1] The paper's key claim is the larger gains on the intervention subset, but the sample size and class balance of this subset are not reported. Table 1 shows absolute accuracy and AUC values, but without knowing how many students are in the intervention subgroup and what the base rate of the positive class is, the reported gains (e.g., MaskMLP 0.7704 vs. MLP Indicator 0.7539 in school-split word identification) cannot be properly interpreted. The authors should report the number of samples and positive-class proportion for the intervention subset and compute a significance test for the MaskMLP versus baseline comparison on that subset.

minor comments (9)

[Abstract, Sec. 4.2] The term 'Multi-Layer Perception' should be 'Multi-Layer Perceptron' (also in Sec. 4.2 where 'perception' is used).
[Sec. 5.1] The phrase 'by10%' is missing a space; it should read 'by 10%'.
[Sec. 5.1] The MLP baseline used in the paired t-test is not specified (which missing-data variant: zeros, mean, or indicator?). Please clarify which baseline configuration is compared with MaskMLP.
[Sec. 5.1, Table 2] The definition of the High FRL and Low FRL subgroups in Table 2 is unclear, since the paper states that FRL data are only available at the school level. Please explain how student-level subgroups are constructed and report the sample sizes per subgroup, as some entries (e.g., High FRL intervention group) may be based on very few students.
[Sec. 5.1, Fig. 2] A 'five-class model' is used for the quantile breakdown in Fig. 2, but this model is not described in Sec. 4. Please specify how the five quantile classes are defined and whether this is a separate classification model from the main binary task.
[Sec. 4.2] The description of the masking process should explicitly note that the 'original input' also contains -1 for missing values, so both the original and masked embeddings are computed from partially observed inputs. The current wording could be misinterpreted as using complete data for the original input.
[Sec. 2, References] Some cited references appear unrelated to the surrounding text, e.g., [35] on assistive navigation and [55] on Turing Award winners. Please check that all citations are relevant and properly integrated.
[Sec. 6] The checklist states that no error bars across seeds are reported, but this limitation is not mentioned in Sec. 6. Add a sentence acknowledging this limitation and its implications for the robustness of the results.
[Supplementary Table 4] The loss ablation table uses checkmarks but does not label the columns with the loss names. Add column headers or a legend so the reader can map the checkmarks to the four loss functions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the core claim is an empirical benchmark comparison, and the cohort-relative label is a task definition rather than a fitted parameter or a reused prediction.

full rationale

The paper's central claims are empirical: MaskMLP, a masked-input MLP pretraining objective, is compared with tabular baselines on a newly released ECRI dataset. No step of this comparison derives a result from its own input by construction. The pretraining loss (Eq. 1) is a cosine embedding loss between embeddings of the original and randomly masked versions of the same feature vector; it does not use labels, and the subsequent fine-tuning uses the defined binary outcome. The binary outcome is defined in Sec. 3.1 and 4.1 as whether a student's gain exceeds the control group's average gain. This is a label-construction choice, not a fitted model parameter; the control-group mean is not estimated by the model and is not a function of MaskMLP's predictions, so the benchmark does not reduce to a fit renamed as a prediction. The paper's self-citations (e.g., ECRI intervention studies [6, 24, 47, 56]) are used to motivate the intervention context and dataset provenance, not as the evidence for the MaskMLP accuracy gains, so they are not load-bearing. The label's dependence on end-of-year control-group information is a genuine deployment and construct-validity concern, and it could become leakage if test-fold outcomes defined the threshold, which the paper does not specify; however, that is a correctness and task-design limitation, not a circular derivation under the criteria used here. Section 6's limitation statement concerns generalizability and bias, not circularity. No uniqueness theorem or ansatz is imported from the authors' prior work to force the choice of MaskMLP. Accordingly, no circular step is identified.

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

The central claim rests on the validity of the reading assessments, the cohort-relative label definition, the transfer of self-supervised representations, and the assumption that the complete-case evaluation set is representative. The only free parameters are architectural and masking hyperparameters selected via ablations on the benchmark itself; no physical or causal entities are introduced.

free parameters (4)

Mask ratio for pre-training = 0.25
Randomly mask 25% of observed variables during pre-training; chosen by the authors without a separate validation-based selection procedure reported.
MLP hidden size = 64
Three-layer MLP with hidden size 64 reported as best in ablations; hyperparameter selection appears to use the same dataset.
Number of MLP layers = 3
Reported as best in ablations; chosen via model selection on the dataset.
Pre-training loss choice = Cosine (with MSE in some settings)
Loss function ablation in supplementary shows Cosine or Cosine+MSE works best; the choice is made based on the same benchmark data.

assumptions (5)

domain assumption The reading assessment instruments (DIBELS, SAT, WRMT, TKP, RCMIS) are valid and reliable measures of early reading skill.
The paper relies on these instruments as ground-truth inputs and outcomes, citing prior validation literature but not re-establishing validity.
domain assumption The binary outcome (improvement above the control group average) is a meaningful target for identifying students who need additional support.
The label is defined relative to the cohort's control-group mean gain; this makes the task depend on the specific control group distribution and may not transfer across settings.
domain assumption Marking missing values with -1 and training a cosine-embedding objective on masked inputs yields representations useful for the downstream classification.
The effectiveness of MaskMLP pre-training is assumed to transfer to the fine-tuning task; this is an empirical claim validated only on this dataset.
standard math The five-fold group split with grouping by student and school ID produces independent training and test sets.
Group k-fold prevents direct student/school overlap, but students within the same school are correlated, so school-split is the cleaner generalization test; student-split may still leak school-level effects.
domain assumption Complete-case analysis for fine-tuning and evaluation does not systematically bias the comparison.
The paper excludes samples with missing data from fine-tuning and evaluation while keeping them in pre-training; if missingness is related to student outcomes (e.g., absent students are lower-performing), the evaluation set is unrepresentative.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Scalable Early Childhood Reading Performance Prediction." pith.science (2026). https://pith.science/paper/KELXOXX2

@misc{pith2026241210401,
  author       = {Pith},
  title        = {Pith review of: Scalable Early Childhood Reading Performance Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KELXOXX2}},
  note         = {Machine review of arXiv:2412.10401}
}

read the original abstract

Models for student reading performance can empower educators and institutions to proactively identify at-risk students, thereby enabling early and tailored instructional interventions. However, there are no suitable publicly available educational datasets for modeling and predicting future reading performance. In this work, we introduce the Enhanced Core Reading Instruction ECRI dataset, a novel large-scale longitudinal tabular dataset collected across 44 schools with 6,916 students and 172 teachers. We leverage the dataset to empirically evaluate the ability of state-of-the-art machine learning models to recognize early childhood educational patterns in multivariate and partial measurements. Specifically, we demonstrate a simple self-supervised strategy in which a Multi-Layer Perception (MLP) network is pre-trained over masked inputs to outperform several strong baselines while generalizing over diverse educational settings. To facilitate future developments in precise modeling and responsible use of models for individualized and early intervention strategies, our data and code are available at https://ecri-data.github.io/.

Figures

Figures reproduced from arXiv: 2412.10401 by the authors.

**Figure 1.** Self-Supervised MLP Pre-Training. We randomly mask parts of the input variables, i.e., as missing values, and train the model using a loss derived from both the original and masked input to a common feature extractor (we employ a cosine embedding loss to enforce similarity among the two embeddings). We consider the problem of learning to classify student progress from observed or partially observed variables x ∈ … view at source ↗

**Figure 2.** Breakdown Results Across Quantile Groups on ECRI. A five-class model accuracy breakdown with classes defined over five quantiles of student performance by improvement amount from the first assessment, from large regression, slight regression, no change, slight improvement, and large improvement. affects students’ reading skills (this underlines the significant role of such interventions). Our proposed MaskMLP method… view at source ↗

**Figure 3.** Visualization of t-SNE-based Embedding and Student Profile Analysis. The visualization uses embeddings derived from the MLP (left) and MaskMLP (right) models for the word identification task, with negative samples shaded in gray and positive samples shaded in green. The pre-training step in MaskMLP results in an embedding with greater separation among student profiles. model exhibits lower success rates for student… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Feature Importance Analysis. We show feature importance by plotting the decrease in model accuracy on word identification classification after dropping each input variable, one at a time. All literacy assessment measures used for input are obtained at the start of the…

**Figure 5.** Figure 5: Race and Ethnicity Distribution across Different Socio-Economic Groups. Left: Distribution for the five schools with the highest percentage of students receiving free or reducedprice lunch. Right: Distribution for the five schools with the lowest percentage of student…

**Figure 6.** Figure 6: Visulization of School-level Features. Left: the percentage of schools providing free or reduced-price lunch. Right: student demographics, averaged over schools. per classroom per day for instruction, plus 1 hour per classroom per week for preparation), along with $250…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 70 canonical work pages

[1]

Agrawal and H

H. Agrawal and H. Mavani. Student performance prediction using machine learning. Int. J. Eng. Res. Technol., 2015

work page 2015
[2]

Ahmed, R

Y . Ahmed, R. Wagner, and D. Lopez. Longitudinal study on reading and writing at the word, sentence, and text levels. LDbase, 2022

work page 2022
[3]

S. Ö. Arik and T. Pfister. Tabnet: Attentive interpretable tabular learning. In AAAI, 2021

work page 2021
[4]

Bahri, H

D. Bahri, H. Jiang, Y . Tay, and D. Metzler. Scarf: Self-supervised contrastive learning using random feature corruption. In ICLR, 2021

work page 2021
[5]

Baker and D

L. Baker and D. Scher. Motivations for reading scale. Read. Psychol., 2002

work page 2002
[6]

S. K. Baker, H. Fien, and D. L. Baker. Robust reading instruction in the early grades: Conceptual and practical issues in the integration and evaluation of tier 1 and tier 2 instructional supports. FOEC, 2010

work page 2010
[7]

S. K. Baker, K. Smolkowski, E. A. Chaparro, J. L. Smith, and H. Fien. Using regression discontinuity to test the impact of a tier 2 reading intervention in first grade. J. Res. Educ. Eff., 2015

work page 2015
[8]

C. M. Bishop and N. M. Nasrabadi. Pattern recognition and machine learning. 2006

work page 2006

Show all 74 references

[9]

R. M. Bruno and S. C. Walker. Comprehensive test of phonological processing (ctopp). SAGE, 1999

1999
[10]

R. D. Camino, C. A. Hammerschmidt, and R. State. Improving missing data imputation with deep generative models. arXiv, 2019. 10

2019
[11]

Cantwell

M. Cantwell. Cantwell, moran introduces bill to boost AI education. PressRelease, 2024

2024
[12]

Celik, M

I. Celik, M. Dindar, H. Muukkonen, and S. Järvelä. The promises and challenges of artificial intelligence for teachers: A systematic review of research. TechTrends, 2022

2022
[13]

Chen and C

T. Chen and C. Guestrin. Xgboost: A scalable tree boosting system. In KDD, 2016

2016
[14]

Y .-W. Chu, S. Hosseinalipour, E. Tenorio, L. Cruz, K. Douglas, A. Lan, and C. Brinton. Mitigating biases in student performance prediction via attention-based personalized federated learning. In CIKM, 2022

2022
[15]

N. H. Clemens, E. Oslund, O.-m. Kwok, M. Fogarty, D. Simmons, and J. L. Davis. Skill moderators of the effects of a reading comprehension intervention. EC, 2019

2019
[16]

M. D. Coyne, D. B. McCoach, S. Ware, C. R. Austin, S. M. Loftus-Rattan, and D. L. Baker. Rac- ing against the vocabulary gap: Matthew effects in early vocabulary instruction and intervention. EC, 2019

2019
[17]

J. Demšar. Statistical comparisons of classifiers over multiple data sets. JMLR, 2006

2006
[18]

Devlin, M.-W

J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv, 2018

2018
[19]

T. G. Dietterich. Approximate statistical tests for comparing supervised classification learning algorithms. Neural Comput., 1998

1998
[20]

Djambic, M

G. Djambic, M. Krajcar, and D. Bele. Machine learning model for early detection of higher education students that need additional attention in introductory programming courses. Int. J. Digit. Enterp. Technol., 2016

2016
[21]

Doabler and N

C. Doabler and N. Nelson-Walker. Ratings of classroom management and instructional support. CTL, UO, 2009

2009
[22]

Doersch and A

C. Doersch and A. Zisserman. Multi-task self-supervised visual learning. In ICCV, 2017

2017
[23]

Enders, S

C. Enders, S. Dietz, M. Montague, and J. Dixon. Modern alternatives for dealing with missing data in special education research. In J. Res. Methodol. 2006

2006
[24]

H. Fien, N. J. Nelson, K. Smolkowski, D. Kosty, M. Pilger, S. K. Baker, and J. L. M. Smith. A conceptual replication study of the enhanced core reading instruction mtss-reading model. EC, 2021

2021
[25]

Fuchs, D

D. Fuchs, D. M. Kearns, L. S. Fuchs, A. M. Elleman, J. K. Gilbert, S. Patton, P. Peng, and D. L. Compton. Using moderator analysis to identify the first-grade children who benefit more and less from a reading comprehension program: A step toward aptitude-by-treatment interacti...

2019
[26]

Gebru, J

T. Gebru, J. Morgenstern, B. Vecchione, J. W. Vaughan, H. Wallach, H. D. Iii, and K. Crawford. Datasheets for datasets. Commun. ACM, 2021

2021
[27]

Giannakas, C

F. Giannakas, C. Troussas, I. V oyiatzis, and C. Sgouropoulou. A deep learning classification framework for early prediction of team-based academic performance. Appl. Soft Comput., 2021

2021
[28]

Gomez, Y

L. Gomez, Y . Patel, M. Rusinol, D. Karatzas, and C. Jawahar. Self-supervised learning of visual features through embedding images into text topic spaces. In CVPR, 2017

2017
[29]

R. H. Good, J. Gruba, and R. A. Kaminski. Best practices in using dynamic indicators of basic early literacy skills (dibels) in an outcomes-driven model. NASP, 2002

2002
[30]

R. H. Good and R. A. Kaminski. Dibels oral reading fluency passages for first through third grades. Tech. Rep., 2002

2002
[31]

J. Han, J. Pei, and H. Tong. Data mining: concepts and techniques. 2022

2022
[32]

Stanford achievement test (10th ed.)

Harcourt Educational Measurement. Stanford achievement test (10th ed.). Harcourt, 2002

2002
[33]

S. Haykin. Neural networks: a comprehensive foundation. 1998

1998
[34]

K. He, X. Chen, S. Xie, Y . Li, P. Dollár, and R. Girshick. Masked autoencoders are scalable vision learners. In CVPR, 2022

2022
[35]

Huang, Z

Z. Huang, Z. Shangguan, J. Zhang, G. Bar, M. Boyd, and E. Ohn-Bar. Assister: Assistive navigation via conditional instruction generation. In ECCV, 2022

2022
[36]

Hussain and M

S. Hussain and M. Q. Khan. Student-performulator: Predicting students’ academic performance at secondary and intermediate level using machine learning. Ann. Data Sci., 2023. 11

2023
[37]

L. Irwin. Early child development: A powerful equalizer. CSDH, 2007

2007
[38]

Kabakchieva

D. Kabakchieva. Predicting student performance by using data mining methods for classification. CIT, 2013

2013
[39]

R. A. Kaminski and R. H. Good III. Toward a technology for assessing basic early literacy skills. SPR, 1996

1996
[40]

G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.-Y . Liu. Lightgbm: A highly efficient gradient boosting decision tree. NeurIPS, 2017

2017
[41]

Kolesnikov, X

A. Kolesnikov, X. Zhai, and L. Beyer. Revisiting self-supervised visual representation learning. In CVPR, 2019

2019
[42]

LeCun, Y

Y . LeCun, Y . Bengio, and G. Hinton. Deep learning.Nature, 2015

2015
[43]

M. Li, R. Xu, S. Wang, L. Zhou, X. Lin, C. Zhu, M. Zeng, H. Ji, and S.-F. Chang. Clip-event: Connecting text and images with event structures. In CVPR, 2022

2022
[44]

Z. C. Lipton, D. C. Kale, R. Wetzel, et al. Modeling missing data in clinical time series with RNNs. MLHC, 2016

2016
[45]

L. Moats. Knowledge foundations for teaching reading and spelling. RW, 2009

2009
[46]

National assessment of educational progress report card: 2022 naep reading assessment

National Assessment of Educational Progress. National assessment of educational progress report card: 2022 naep reading assessment. Online, 2022

2022
[47]

N. J. Nelson-Walker, H. Fien, D. B. Kosty, K. Smolkowski, J. L. M. Smith, and S. K. Baker. Eval- uating the effects of a systemic intervention on first-grade teachers’ explicit reading instruction. LDQ, 2013

2013
[48]

K. Niu, X. Cao, and Y . Yu. Explainable student performance prediction with personalized attention for explaining why a student fails. arXiv, 2021

2021
[49]

Norouzi and A

N. Norouzi and A. Mazaheri. Context-aware analysis of group submissions for group anomaly detection and performance prediction. In AAAI, 2023

2023
[50]

C.-Y . J. Peng, M. Harwell, S.-M. Liou, L. H. Ehman, et al. Advances in missing data methods and implications for educational research. RDA, 2006

2006
[51]

T. Peng, Y . Liang, W. Wu, J. Ren, Z. Pengrui, and Y . Pu. Clgt: A graph transformer for student performance prediction in collaborative learning. In AAAI, 2023

2023
[52]

J. L. Peugh and C. K. Enders. Missing data in educational research: A review of reporting practices and suggestions for improvement. Rev. Educ. Res., 2004

2004
[53]

Radford, J

A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. In ICML, 2021

2021
[54]

Shahbazian and I

R. Shahbazian and I. Trubitsyna. Degain: Generative-adversarial-network-based missing data imputation. Information, 2022

2022
[55]

Shangguan, Z

Z. Shangguan, Z. Zheng, and J. Luo. What kind of person wins the turing award? arXiv:2104.05636, 2021

2021 arXiv
[56]

J. L. M. Smith, N. J. Nelson, H. Fien, K. Smolkowski, D. Kosty, and S. K. Baker. Examining the efficacy of a multitiered intervention for at-risk readers in grade 1. ESJ, 2016

2016
[57]

Smolkowski and B

K. Smolkowski and B. Gunn. Reliability and validity of the classroom observations of student– teacher interactions (costi) for kindergarten reading instruction. ECRQ, 2012

2012
[58]

Van der Maaten and G

L. Van der Maaten and G. Hinton. Visualizing data using t-SNE. JMLR, 2008

2008
[59]

Vaughn, G

S. Vaughn, G. Roberts, P. Capin, J. Miciak, E. Cho, and J. M. Fletcher. How initial word reading and language skills affect reading comprehension outcomes for students with reading difficulties. EC, 2019

2019
[60]

R. W. Woodcock. Woodcock reading mastery tests, revised. AGS, 1998

1998
[61]

J. Xu, K. H. Moon, and M. Van Der Schaar. A machine learning approach for tracking and predicting student performance in degree programs. J. Sel. Top. Signal Process, 2017

2017
[62]

Xu and F

W. Xu and F. Ouyang. The application of ai technologies in stem education: a systematic review from 2011 to 2021. Int. j. STEM educ. , 2022. 12

2011
[63]

W. M. Yen, A. R. Fitzpatrick, and R. Brennan. Educational measurement. PRG, 2006

2006
[64]

J. Yoon, J. Jordon, and M. Schaar. Gain: Missing data imputation using generative adversarial nets. In ICML, 2018

2018
[65]

J. Yoon, Y . Zhang, J. Jordon, and M. van der Schaar. Vime: Extending the success of self-and semi-supervised learning to tabular domain. NeurIPS, 2020

2020
[66]

J. You, X. Ma, Y . Ding, M. J. Kochenderfer, and J. Leskovec. Handling missing data with graph representation learning. NeurIPS, 2020

2020
[67]

Zhang, M

J. Zhang, M. Zheng, M. Boyd, and E. Ohn-Bar. X-world: Accessibility, vision, and autonomy meet. In CVPR, 2021

2021
[68]

Zhang, Y

Y . Zhang, Y . Yun, R. An, J. Cui, H. Dai, and X. Shang. Educational data mining techniques for student performance prediction: method review and comparison analysis. Front. Psychol., 2021

2021
[69]

Zheng and N

S. Zheng and N. Charoenphakdee. Diffusion models for missing value imputation in tabular data. arXiv, 2022. 8 Checklist

2022
[70]

For all authors... (a) Do the main claims made in the abstract and introduction accurately reflect the paper’s contributions and scope? [Yes] (b) Did you describe the limitations of your work? [Yes] See Sec. 6. (c) Did you discuss any potential negative societal impacts of you...
[71]

(a) Did you state the full set of assumptions of all theoretical results? [N/A] (b) Did you include complete proofs of all theoretical results? [N/A]

If you are including theoretical results... (a) Did you state the full set of assumptions of all theoretical results? [N/A] (b) Did you include complete proofs of all theoretical results? [N/A]
[72]

for benchmarks)

If you ran experiments (e.g. for benchmarks)... (a) Did you include the code, data, and instructions needed to reproduce the main experi- mental results (either in the supplemental material or as a URL)? [Yes] (b) Did you specify all the training details (e.g., data splits, hy...
[73]

(a) If your work uses existing assets, did you cite the creators? [Yes] We cite all creators and authors related to our data and methods where necessary

If you are using existing assets (e.g., code, data, models) or curating/releasing new assets... (a) If your work uses existing assets, did you cite the creators? [Yes] We cite all creators and authors related to our data and methods where necessary. (b) Did you mention the lic...
[74]

If you used crowdsourcing or conducted research with human subjects... 13 (a) Did you include the full text of instructions given to participants and screenshots, if applicable? [N/A] (b) Did you describe any potential participant risks, with links to Institutional Review Boar...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

Agrawal and H

H. Agrawal and H. Mavani. Student performance prediction using machine learning. Int. J. Eng. Res. Technol., 2015

work page 2015

[2] [2]

Ahmed, R

Y . Ahmed, R. Wagner, and D. Lopez. Longitudinal study on reading and writing at the word, sentence, and text levels. LDbase, 2022

work page 2022

[3] [3]

S. Ö. Arik and T. Pfister. Tabnet: Attentive interpretable tabular learning. In AAAI, 2021

work page 2021

[4] [4]

Bahri, H

D. Bahri, H. Jiang, Y . Tay, and D. Metzler. Scarf: Self-supervised contrastive learning using random feature corruption. In ICLR, 2021

work page 2021

[5] [5]

Baker and D

L. Baker and D. Scher. Motivations for reading scale. Read. Psychol., 2002

work page 2002

[6] [6]

S. K. Baker, H. Fien, and D. L. Baker. Robust reading instruction in the early grades: Conceptual and practical issues in the integration and evaluation of tier 1 and tier 2 instructional supports. FOEC, 2010

work page 2010

[7] [7]

S. K. Baker, K. Smolkowski, E. A. Chaparro, J. L. Smith, and H. Fien. Using regression discontinuity to test the impact of a tier 2 reading intervention in first grade. J. Res. Educ. Eff., 2015

work page 2015

[8] [8]

C. M. Bishop and N. M. Nasrabadi. Pattern recognition and machine learning. 2006

work page 2006