Pith. sign in

REVIEW 4 major objections 4 minor 50 references

MultiTab: A Comprehensive Benchmark Suite for Multi-Dimensional Evaluation in Tabular Domains

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Tabular model performance flips with data regime, and average benchmarks hide it.

desk verdict A genuinely reusable benchmark artifact whose headline regime findings are plausible but currently rest on a fragile relative-error metric and some internal inconsistencies. read the letter →

arxiv 2505.14312 v1 pith:LFX77TMS submitted 2025-05-20 cs.LG cs.AI

Kyungeun Lee , Moonjung Eo , Hye-Seung Cho , Dongmin Kim , Ye Seul Sim , Seoyoon Kim , Min-Kook Suh , Woohyung Lim This is my paper
classification cs.LGcs.AI
keywords tabulardatabenchmarkregimesnormalizedpredictiveerrorinductivebiassample-levelsimilarityfeatureinteractionmodelselection
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 argues that how well a tabular model performs depends strongly on the statistical shape of the dataset, and that average-case benchmarks hide these conditional strengths. It builds MultiTab, a suite of 196 publicly available datasets evaluated with 13 models, and groups the datasets along seven data-characteristic axes such as sample size, label imbalance, and feature correlation. The central finding is a stable regime sensitivity: models built on sample-level similarity win on large or highly correlated datasets, feature-attention models win on weakly correlated features, and gradient-boosted trees remain strong on regression and low-imbalance tasks. If the finding is right, model selection and architecture design should be data-regime-aware, and benchmark reporting should include conditional results rather than a single global ranking.

What carries the argument

MultiTab's sub-categorization framework partitions 196 datasets along seven axes, each measured by quantitative statistics: task type, sample size, feature heterogeneity (proportion and cardinality of categorical features), feature-to-sample ratio, label imbalance (entropy ratio, skewness, imbalance factor), function irregularity (a frequency-based high-frequency energy ratio), and feature interaction (Frobenius norm of correlation and minimum eigenvalue of the covariance matrix). Models are grouped into six inductive-bias families. The primary metric is normalized predictive error: within each dataset-split pair, each model's error is linearly rescaled between the best and worst errors to a value in [0,1], then averaged across splits, which is meant to make relative performance comparable across datasets of very different difficulty.

What would settle it

Recompute the per-regime rankings using raw errors or average ranks instead of normalized scores, and check whether the reported patterns, NN-Sample best on large and correlated data and NN-Feature best on weakly correlated data, persist under both metrics and across splits with different dataset compositions per sub-category; if the ordering flips, the central claim is an artifact of normalization.

Watch

Extended reading notes

Core claim

The central claim is that model performance on tabular data is highly sensitive to data regimes, and that the regime-specific rankings are systematic rather than noise. Specifically, the paper reports that models relying on sample-level similarity excel on datasets with large sample sizes or high inter-feature correlation, while models encoding inter-feature dependencies perform best when features are weakly correlated; gradient-boosted trees dominate regression and low-imbalance settings. The paper also finds that inductive biases do not always behave as intended, citing cases where a model designed for both feature and sample dependencies underperforms specialized architectures, and where a tree-based model known for robustness degrades sharply under function irregularity. These patterns support the conclusion that average-case rankings are insufficient for tabular model evaluation.

Load-bearing premise

The main load-bearing assumption is that rescaling each model's error to a 0-to-1 range within every dataset-split and then averaging these relative scores makes errors comparable across datasets of very different difficulty, so the reported regime differences reflect true model strengths rather than artifacts of the normalization scale.

Editorial extensions

If this is right

  • Practitioners can choose a model based on dataset statistics rather than relying on global benchmark rankings.
  • Benchmark suites should report performance across data regimes as a standard practice, not only aggregate averages.
  • The identified weak spots, such as neural models degrading under skewed regression targets and all models struggling on highly irregular functions, point to concrete gaps for architectural improvement.
  • Models that exploit sample-level similarity are a reasonable default for large, highly correlated datasets, whereas feature-attention models suit weakly correlated, high-dimensional data.
  • Inductive-bias labels alone do not predict success; implementation-level behavior within a model family can diverge sharply across regimes.

Reading between the lines

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

  • The normalized-error protocol may reward models that are consistently close to the best on each dataset, so the regime patterns could partly reflect compression of error scales; checking the same claims with raw ranks or fold-level statistical tests would sharpen them.
  • The dataset axes are not independent, so the attribution of performance differences to a single characteristic such as sample size may be confounded with feature-to-sample ratio or correlation structure; synthetic datasets that vary one axis at a time could test the claimed mechanisms.
  • The reported TabPFN comparison suggests pretrained tabular models may occupy a niche in imbalanced small-scale classification, which is a testable prediction for tabular foundation models as they scale.
  • The 'inductive bias does not behave as intended' finding implies that reporting per-family averages can be misleading; future conditional evaluations should report per-model results alongside family-level summaries.
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

4 major / 4 minor

Summary. The paper introduces MultiTab, a benchmark suite and evaluation framework for tabular learning that moves beyond average-case rankings. It collects 196 publicly available datasets, evaluates 13 models spanning classical methods, GBDTs, and neural networks with different inductive biases (sample-level similarity, feature-level attention, both, or neither), and partitions datasets along seven axes (task type, sample size, feature heterogeneity, feature-to-sample ratio, label imbalance, function irregularity, and feature interaction) using heuristic thresholds. The primary metric is a per-dataset-split min-max normalized predictive error, averaged across dataset-split pairs. The central empirical claim is that model performance is regime-dependent: for example, NN-Sample models excel on large-sample or high-correlation datasets, while NN-Feature models excel when features are weakly correlated, and that these patterns are invisible under aggregate evaluation. The paper also reports Spearman correlations between dataset statistics and model error, a comparison with TabPFN on 42 eligible datasets, and extensive hyperparameter optimization logs.

Significance. If the results are robust, MultiTab would be a valuable community resource: it is large in scale (196 datasets, 13 models, 100 TPE trials per model-dataset pair with per-fold optimization), it is released with optimization logs, and it targets a real gap in the literature by proposing conditional, regime-aware evaluation instead of single-number averages. The paper also deserves credit for attempting to connect architectural inductive biases to dataset statistics in a falsifiable way, and for including regression tasks and multiple complementary metrics (entropy ratio, skewness, imbalance factor) for label imbalance. However, the central claims currently rest on a normalized-error metric whose validity is not demonstrated, and on split-level aggregation that makes the reported confidence intervals unreliable; these issues must be addressed before the regime-level conclusions can be accepted.

major comments (4)
  1. [§3.4, Appendix B.3] The primary metric is the per-dataset-split min-max normalized error, \hat e_{m,d} = (e_{m,d} - e_min_d)/(e_max_d - e_min_d), computed over the 13 models evaluated on that split. This is a relative, rank-like score: a model that is best by a tiny margin receives 0 and a model that is worst by a tiny margin receives 1, so averaging these scores across heterogeneous datasets can amplify negligible raw-error differences into the "excels" claims in §4.2. Since the raw log loss/RMSE and average-rank results in Appendix E.1 are not used to validate the regime-level conclusions, the central findings currently rest on a metric that may not reflect meaningful absolute improvements. I request a sensitivity analysis: report the raw-error margins for the highlighted comparisons (e.g., ModernNCA vs XGBoost in Table 2), and re-run the key regime analyses with raw log loss/RMSE or average rank as the primary outcome.
  2. [§3.3–3.4, Table 2, Figure 1] Aggregation is performed over dataset-split pairs, not datasets: each dataset contributes 10 folds (or 3 folds for large datasets), and fold-level scores from the same dataset are correlated because they share samples and validation strategies. The 95% confidence intervals and "statistically comparable" statements in Table 2 and Figure 1 are therefore computed over non-independent units and will overstate significance, while also weighting each small dataset more heavily than each large dataset. A dataset-level analysis (for example, averaging over folds first and then using paired tests or bootstrap over datasets) is needed to support the significance claims.
  3. [Table 1 vs Appendix C] There are direct inconsistencies in the reported sub-category definitions. Table 1 states function-irregularity thresholds of 0.25 (Regular) and 0.95 (Irregular) with 38 and 43 datasets, while Appendix C.6 uses 0.3 and 0.7 with 58 and 43 datasets. For entropy ratio, Table 1 reports Balanced >0.7 with 43 datasets and Imbalanced <0.3 with 61, whereas Appendix C.5.1 reports 61 and 43, respectively. These differences change the dataset composition of the regimes and make it impossible to reproduce Table 2 and Figure 2 from the appendix alone. Please harmonize the thresholds and counts, and state which definitions generated the reported results.
  4. [Appendix B.1, §3.1] The preprocessing section excludes datasets with categorical features having more than 1,000 unique values "due to frequent training failures across models." The number and identity of excluded datasets are not reported, and no sensitivity analysis is given. Because the excluded datasets are likely concentrated in exactly the high-cardinality categorical regime analyzed in §4.2.3, this exclusion could bias the very comparison it is used to support. Please report the excluded datasets and assess whether the categorical-cardinality conclusions change when they are handled through a different encoding or excluded with a documented justification.
minor comments (4)
  1. [Figure 2] The caption states "Blue (negative) indicates worse-than-average performance; red (positive) indicates better-than-average," but a negative deviation from the overall mean error should indicate better-than-average performance. Please clarify the color/sign convention and make the text in §4.2.3 consistent with it.
  2. [Appendix D.5, Tables 12 and 13] There are typos in the hyperparameter tables: "d_embedings" and "frequency_sclae" in the ModernNCA table should be "d_embeddings" and "frequency_scale." The text in §D.5.2 also says "d_embedings" in the paragraph preceding Table 12.
  3. [Appendix E.4] The sentence "where as ModernNCA often achieves larger improvements" should read "whereas." Also, Appendix E.3 contains a reference to "Figure ??" that should be replaced with the actual figure number.
  4. [Appendix B] The statement that "All datasets are distributed under the CC-BY license" is likely too strong for OpenML datasets, which have heterogeneous licenses. Please verify the license of every dataset or soften the claim to describe the verified subset.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: MultiTab's regime-level findings are empirical benchmark observations, not derivations that reduce to the evaluation metric or to self-citations.

full rationale

MultiTab is an empirical evaluation suite, not a derivation of a predicted quantity from a fitted model. The central claims (e.g., NN-Sample excels on large-sample/high-correlation datasets; NN-Feature excels under weak correlation) are statements about measured normalized predictive error across defined dataset sub-categories. The normalized error metric in Section 3.4 and Appendix B.3 rescales raw errors within each dataset-split pair, but this is an explicit aggregation choice; the regime conclusions are not implied by the formula alone and could have come out differently. No parameter is fitted to a subset of data and then renamed as a prediction, and no uniqueness theorem or ansatz is imported from prior work to force a choice. The only self-citations (refs [18,19]) are used for dataset-selection references and California-housing provenance in Appendix B; they do not carry the load of any empirical finding. Concerns about the normalized metric's comparability across splits and about confidence intervals computed over non-independent folds are methodological validity issues, not circularity under the defined analysis. The paper is therefore self-contained against external benchmarks for the purposes of this pass.

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

The analysis rests on hand-chosen thresholds for each data axis, a relative error normalization, and the assumption that the selected public datasets are representative. No new theoretical entities are invented, but the thresholds and metric choices are free parameters of the evaluation.

free parameters (5)
  • Sample size regime thresholds = small < 1000, large > 10000
    Chosen empirically from histogram distributions (Appendix C.2) and used to define the regimes in the central analysis.
  • Categorical proportion thresholds = few < 20%, many > 60%
    Chosen from threshold conventions in prior work; used to define feature heterogeneity regimes (Appendix C.3.1).
  • Entropy ratio thresholds = balanced > 0.7, imbalanced < 0.3
    Table 1 vs Appendix C.5.1 show swapped counts; thresholds are heuristic and the inconsistency is unresolved.
  • Function irregularity thresholds = regular < 0.25/0.3, irregular > 0.95/0.7 (paper is inconsistent)
    Main text Table 1 and Appendix C.6 use different cutoffs, which is a reproducibility gap.
  • Feature interaction thresholds = Frobenius norm: correlated > 0.03, uncorrelated < 0.005; min eigenvalue: correlated < 0.002, uncorrelated > 0.1
    Heuristic cutoffs used to partition datasets along the feature-interaction axis.
assumptions (4)
  • domain assumption OpenML and scikit-learn dataset collections are representative of real-world tabular prediction tasks
    All findings are drawn from 196 datasets sourced from these repositories (Section 3.1).
  • domain assumption The chosen statistics (entropy ratio, Frobenius norm of correlation, high-frequency energy ratio, etc.) accurately measure the intended data characteristics
    The sub-category axes are defined by these statistics (Section 3.5). If a metric does not capture its named property, the regime-specific findings do not mean what the prose says.
  • ad hoc to paper Min-max normalized error is comparable across datasets and splits
    The primary metric in Section 3.4 rescales errors within each dataset-split; averaging these normalized scores presumes cross-dataset comparability.
  • domain assumption 100 TPE optimization trials provide fair and sufficient tuning for every model-dataset pair
    Training protocol in Section 3.3 fixes the trial budget; if some architectures need more trials, comparisons are biased.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MultiTab: A Comprehensive Benchmark Suite for Multi-Dimensional Evaluation in Tabular Domains." pith.science (2026). https://pith.science/paper/LFX77TMS

@misc{pith2026250514312,
  author       = {Pith},
  title        = {Pith review of: MultiTab: A Comprehensive Benchmark Suite for Multi-Dimensional Evaluation in Tabular Domains},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LFX77TMS}},
  note         = {Machine review of arXiv:2505.14312}
}
read the original abstract

Despite the widespread use of tabular data in real-world applications, most benchmarks rely on average-case metrics, which fail to reveal how model behavior varies across diverse data regimes. To address this, we propose MultiTab, a benchmark suite and evaluation framework for multi-dimensional, data-aware analysis of tabular learning algorithms. Rather than comparing models only in aggregate, MultiTab categorizes 196 publicly available datasets along key data characteristics, including sample size, label imbalance, and feature interaction, and evaluates 13 representative models spanning a range of inductive biases. Our analysis shows that model performance is highly sensitive to such regimes: for example, models using sample-level similarity excel on datasets with large sample sizes or high inter-feature correlation, while models encoding inter-feature dependencies perform best with weakly correlated features. These findings reveal that inductive biases do not always behave as intended, and that regime-aware evaluation is essential for understanding and improving model behavior. MultiTab enables more principled model design and offers practical guidance for selecting models tailored to specific data characteristics. All datasets, code, and optimization logs are publicly available at https://huggingface.co/datasets/LGAI-DILab/Multitab.

Figures

Figures reproduced from arXiv: 2505.14312 by the authors.

Figure 1
Figure 1. Average normalized predictive error across 24 sub-categories for four model classes: GBDTs, NN￾Simple, NN-Sample, and NN-Feature. Lower values indicate better performance. Error bars represent 95% confidence intervals. Unlike overall averages, model rankings vary substantially across different data regimes, highlighting the importance of conditional evaluation. data regimes. Second, several axes are measured using c… view at source ↗
Figure 2
Figure 2. Deviation from overall average error across 24 sub-categories. Each cell shows the difference between a model’s average error in a given sub-category and its overall mean across all datasets. Blue (negative) indicates worse-than-average performance; red (positive) indicates better-than-average. This highlights model-specific strengths and weaknesses relative to their overall behavior. Taken together, these results s… view at source ↗
Figure 3
Figure 3. Red cells denote performance degradation [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Distribution of datasets in our benchmark suite. Each point represents a dataset, plotted by sample size (x-axis) and feature dimensionality (y-axis), both on logarithmic scale. Colors indicate task types: binary classification, multiclass classification, and regressio…
Figure 5
Figure 5. Figure 5: Detailed histograms for each sub-category 23 [PITH_FULL_IMAGE:figures/full_fig_p023_5.png]
Figure 6
Figure 6. Figure 6: shows the model-class-level results based on accuracy and RMSE, computed over all splits and datasets grouped by each sub-category [PITH_FULL_IMAGE:figures/full_fig_p034_6.png]
Figure 7
Figure 7. Figure 7: Average rank across model families and dataset sub-categories, computed using raw log loss and RMSE. Lower is better. E.2 Comparison between individual algorithms Here we report comparisons among the top-performing individual algorithms: XGBoost, FT￾Transformer, T2G-Fo…
Figure 8
Figure 8. Figure 8: Normalized error of top-performing algorithms across dataset sub-categories. Lower is better [PITH_FULL_IMAGE:figures/full_fig_p035_8.png]
Figure 9
Figure 9. Figure 9: Normalized error of top-performing algorithms based on accuracy for classification tasks. Lower is better [PITH_FULL_IMAGE:figures/full_fig_p035_9.png]
Figure 10
Figure 10. Figure 10: Average rank of top-performing algorithms across sub-categories. Lower is better. E.3 Correlation Between Dataset Statistics and Model Performance To understand continuous trends beyond discrete subgroups, we compute Spearman correlations between dataset statistics an…
Figure 11
Figure 11. Figure 11: Relative performance comparison between TabPFN and ModernNCA across 42 datasets where both models are applicable. The x-axis shows the relative difference in error, with negative values indicating TabPFN wins. TabPFN outperforms ModernNCA on 29 datasets (blue region),…
Figure 12
Figure 12. Figure 12: Average training time per model 37 [PITH_FULL_IMAGE:figures/full_fig_p037_12.png]
Figure 13
Figure 13. Figure 13: Average time required to reach the best hyperparameter [PITH_FULL_IMAGE:figures/full_fig_p038_13.png]
Figure 14
Figure 14. Figure 14: Average trial required to reach the best hyperparameter Most influential hyperparameters. To better understand which hyperparameters most strongly influence model performance, we computed feature importance scores using Optuna’s built-in analysis tools. For each model…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 20 canonical work pages

  1. [1]

    Tabular data: Deep learning is not all you need

    Ravid Shwartz-Ziv and Amitai Armon. Tabular data: Deep learning is not all you need. Information Fusion, 81:84–90, 2022

    2022

  2. [2]

    Deep neural networks and tabular data: A survey.IEEE transactions on neural networks and learning systems, 2022

    Vadim Borisov, Tobias Leemann, Kathrin Seßler, Johannes Haug, Martin Pawelczyk, and Gjergji Kasneci. Deep neural networks and tabular data: A survey.IEEE transactions on neural networks and learning systems, 2022

    2022

  3. [3]

    Trust issues: Uncertainty estimation does not enable reliable ood detection on medical tabular data

    Dennis Ulmer, Lotta Meijerink, and Giovanni Cinà. Trust issues: Uncertainty estimation does not enable reliable ood detection on medical tabular data. InMachine Learning for Health, pages 341–354. PMLR, 2020

    work page 2020

  4. [4]

    Deep learning and the electrocardiogram: review of the current state-of-the-art.EP Europace, 23(8):1179–1191, 2021

    Sulaiman Somani, Adam J Russak, Felix Richter, Shan Zhao, Akhil Vaid, Fayzan Chaudhry, Jessica K De Freitas, Nidhi Naik, Riccardo Miotto, Girish N Nadkarni, et al. Deep learning and the electrocardiogram: review of the current state-of-the-art.EP Europace, 23(8):1179–1191, 2021

    work page 2021

  5. [5]

    Robust cognitive load detection from wrist-band sensors.Computers in Human Behavior Reports, 4:100116, 2021

    Vadim Borisov, Enkelejda Kasneci, and Gjergji Kasneci. Robust cognitive load detection from wrist-band sensors.Computers in Human Behavior Reports, 4:100116, 2021

    work page 2021

  6. [6]

    Sequential deep learning for credit risk monitoring with tabular financial data.arXiv preprint arXiv:2012.15330, 2020

    Jillian M Clements, Di Xu, Nooshin Yousefi, and Dmitry Efimov. Sequential deep learning for credit risk monitoring with tabular financial data.arXiv preprint arXiv:2012.15330, 2020

    arXiv 2012

  7. [7]

    Deepfm: a factorization-machine based neural network for ctr prediction.arXiv preprint arXiv:1703.04247, 2017

    Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. Deepfm: a factorization-machine based neural network for ctr prediction.arXiv preprint arXiv:1703.04247, 2017

    arXiv 2017

  8. [8]

    Deep learning based recommender system: A survey and new perspectives.ACM computing surveys (CSUR), 52(1):1–38, 2019

    Shuai Zhang, Lina Yao, Aixin Sun, and Yi Tay. Deep learning based recommender system: A survey and new perspectives.ACM computing surveys (CSUR), 52(1):1–38, 2019

    work page 2019

Show all 50 references
  1. [9]

    A survey of evolution in predictive models and impacting factors in customer churn.Advances in Data Science and Adaptive Analysis, 9(03):1750007, 2017

    Mehreen Ahmed, Hammad Afzal, Awais Majeed, and Behram Khan. A survey of evolution in predictive models and impacting factors in customer churn.Advances in Data Science and Adaptive Analysis, 9(03):1750007, 2017

    2017

  2. [10]

    A customer churn prediction model based on xgboost and mlp

    Qi Tang, Guoen Xia, Xianquan Zhang, and Feng Long. A customer churn prediction model based on xgboost and mlp. In2020 international conference on computer engineering and application (ICCEA), pages 608–612. IEEE, 2020

    2020

  3. [11]

    Revisiting deep learning models for tabular data.Advances in Neural Information Processing Systems, 34:18932– 18943, 2021

    Yury Gorishniy, Ivan Rubachev, Valentin Khrulkov, and Artem Babenko. Revisiting deep learning models for tabular data.Advances in Neural Information Processing Systems, 34:18932– 18943, 2021

    2021

  4. [12]

    Tabpfn: A transformer that solves small tabular classification problems in a second.arXiv preprint arXiv:2207.01848, 2022

    Noah Hollmann, Samuel Müller, Katharina Eggensperger, and Frank Hutter. Tabpfn: A transformer that solves small tabular classification problems in a second.arXiv preprint arXiv:2207.01848, 2022

    2022 arXiv

  5. [13]

    Hyper- fast: Instant classification for tabular data

    David Bonet, Daniel Mas Montserrat, Xavier Giró-i Nieto, and Alexander G Ioannidis. Hyper- fast: Instant classification for tabular data. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 11114–11123, 2024

    2024

  6. [14]

    Why tabular foundation models should be a research priority.arXiv preprint arXiv:2405.01147, 2024

    Boris Van Breugel and Mihaela Van Der Schaar. Why tabular foundation models should be a research priority.arXiv preprint arXiv:2405.01147, 2024

    2024 arXiv

  7. [15]

    Mixed-type tabular data synthesis with score-based diffusion in latent space.arXiv preprint arXiv:2310.09656, 2023

    Hengrui Zhang, Jiani Zhang, Balasubramaniam Srinivasan, Zhengyuan Shen, Xiao Qin, Christos Faloutsos, Huzefa Rangwala, and George Karypis. Mixed-type tabular data synthesis with score-based diffusion in latent space.arXiv preprint arXiv:2310.09656, 2023

    arXiv

  8. [16]

    D2r2: Diffusion-based repre- sentation with random distance matching for tabular few-shot learning.Advances in Neural Information Processing Systems, 37:36890–36913, 2024

    Ruoxue Liu, Linjiajie Fang, Wenjia Wang, and Bingyi Jing. D2r2: Diffusion-based repre- sentation with random distance matching for tabular few-shot learning.Advances in Neural Information Processing Systems, 37:36890–36913, 2024

    2024

  9. [17]

    Arithmetic feature interaction is necessary for deep tabular learning

    Yi Cheng, Renjun Hu, Haochao Ying, Xing Shi, Jian Wu, and Wei Lin. Arithmetic feature interaction is necessary for deep tabular learning. InProceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 11516–11524, 2024. 10

    2024

  10. [18]

    Binning as a pretext task: Improving self-supervised learning in tabular domains.arXiv preprint arXiv:2405.07414, 2024

    Kyungeun Lee, Ye Seul Sim, Hye-Seung Cho, Moonjung Eo, Suhee Yoon, Sanghyu Yoon, and Woohyung Lim. Binning as a pretext task: Improving self-supervised learning in tabular domains.arXiv preprint arXiv:2405.07414, 2024

    2024 arXiv

  11. [19]

    Representation space augmentation for effective self-supervised learning on tabular data

    Moonjung Eo, Kyungeun Lee, Hye-Seung Cho, Dongmin Kim, Ye Seul Sim, and Woohyung Lim. Representation space augmentation for effective self-supervised learning on tabular data. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 11625–11633, 2025

    2025

  12. [20]

    Why do tree-based models still outperform deep learning on typical tabular data?Advances in neural information processing systems, 35:507–520, 2022

    Léo Grinsztajn, Edouard Oyallon, and Gaël Varoquaux. Why do tree-based models still outperform deep learning on typical tabular data?Advances in neural information processing systems, 35:507–520, 2022

    2022

  13. [21]

    When do neural nets outperform boosted trees on tabular data?Advances in Neural Information Processing Systems, 36, 2024

    Duncan McElfresh, Sujay Khandagale, Jonathan Valverde, Vishak Prasad C, Ganesh Ramakr- ishnan, Micah Goldblum, and Colin White. When do neural nets outperform boosted trees on tabular data?Advances in Neural Information Processing Systems, 36, 2024

    2024

  14. [22]

    Tabrepo: A large scale repository of tabular model evaluations and its automl applications.arXiv preprint arXiv:2311.02971, 2023

    David Salinas and Nick Erickson. Tabrepo: A large scale repository of tabular model evaluations and its automl applications.arXiv preprint arXiv:2311.02971, 2023

    2023 arXiv

  15. [23]

    A closer look at deep learning on tabular data.arXiv preprint arXiv:2407.00956, 2024

    Han-Jia Ye, Si-Yang Liu, Hao-Run Cai, Qi-Le Zhou, and De-Chuan Zhan. A closer look at deep learning on tabular data.arXiv preprint arXiv:2407.00956, 2024

    2024

  16. [24]

    Benchmarking distribution shift in tabular data with tableshift.Advances in Neural Information Processing Systems, 36, 2024

    Josh Gardner, Zoran Popovic, and Ludwig Schmidt. Benchmarking distribution shift in tabular data with tableshift.Advances in Neural Information Processing Systems, 36, 2024

    2024

  17. [25]

    Towards heterogeneous long-tailed learning: Benchmarking, metrics, and toolbox.Advances in Neural Information Processing Systems, 37:73098–73123, 2024

    Haohui Wang, Weijie Guan, Chen Jianpeng, Zi Wang, and Dawei Zhou. Towards heterogeneous long-tailed learning: Benchmarking, metrics, and toolbox.Advances in Neural Information Processing Systems, 37:73098–73123, 2024

    2024

  18. [26]

    On embeddings for numerical features in tabular deep learning.Advances in Neural Information Processing Systems, 35:24991–25004, 2022

    Yury Gorishniy, Ivan Rubachev, and Artem Babenko. On embeddings for numerical features in tabular deep learning.Advances in Neural Information Processing Systems, 35:24991–25004, 2022

    2022

  19. [27]

    T2g-former: organizing tabular features into relation graphs promotes heterogeneous feature interaction

    Jiahuan Yan, Jintai Chen, Yixuan Wu, Danny Z Chen, and Jian Wu. T2g-former: organizing tabular features into relation graphs promotes heterogeneous feature interaction. InProceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 10720–10728, 2023

    2023

  20. [28]

    A data-centric perspective on evaluating machine learning models for tabular data.arXiv preprint arXiv:2407.02112, 2024

    Andrej Tschalzev, Sascha Marton, Stefan Lüdtke, Christian Bartelt, and Heiner Stuckenschmidt. A data-centric perspective on evaluating machine learning models for tabular data.arXiv preprint arXiv:2407.02112, 2024

    2024 arXiv

  21. [29]

    A comprehensive benchmark of machine and deep learning across diverse tabular datasets.arXiv preprint arXiv:2408.14817, 2024

    Assaf Shmuel, Oren Glickman, and Teddy Lazebnik. A comprehensive benchmark of machine and deep learning across diverse tabular datasets.arXiv preprint arXiv:2408.14817, 2024

    2024 arXiv

  22. [30]

    Tabred: Analyzing pit- falls and filling the gaps in tabular deep learning benchmarks.arXiv preprint arXiv:2406.19380, 2024

    Ivan Rubachev, Nikolay Kartashev, Yury Gorishniy, and Artem Babenko. Tabred: Analyzing pit- falls and filling the gaps in tabular deep learning benchmarks.arXiv preprint arXiv:2406.19380, 2024

    2024 arXiv

  23. [31]

    Towards quantifying the effect of datasets for benchmarking: A look at tabular machine learning, 2024

    Ravin Kohli, Matthias Feurer, Katharina Eggensperger, Bernd Bischl, and Frank Hutter. Towards quantifying the effect of datasets for benchmarking: A look at tabular machine learning, 2024

    2024

  24. [32]

    van Rijn, Arlind Kadra, Pieter Gijsbers, Neeratyoy Mallik, Sahithya Ravi, Andreas Mueller, Joaquin Vanschoren, and Frank Hutter

    Matthias Feurer, Jan N. van Rijn, Arlind Kadra, Pieter Gijsbers, Neeratyoy Mallik, Sahithya Ravi, Andreas Mueller, Joaquin Vanschoren, and Frank Hutter. Openml-python: an extensible python api for openml.arXiv, 1911.02490

    1911 arXiv

  25. [33]

    Pedregosa, G

    F. Pedregosa, G. Varoquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V . Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. Scikit-learn: Machine learning in Python.Journal of Machine Learnin...

    2011

  26. [34]

    The choice of scaling technique matters for classification performance.Applied Soft Computing, 133:109924, 2023

    Lucas BV de Amorim, George DC Cavalcanti, and Rafael MO Cruz. The choice of scaling technique matters for classification performance.Applied Soft Computing, 133:109924, 2023. 11

    2023

  27. [35]

    Random forests.Machine learning, 45:5–32, 2001

    Leo Breiman. Random forests.Machine learning, 45:5–32, 2001

    2001

  28. [36]

    Xgboost: A scalable tree boosting system

    Tianqi Chen and Carlos Guestrin. Xgboost: A scalable tree boosting system. InProceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, pages 785–794, 2016

    2016

  29. [37]

    Catboost: unbiased boosting with categorical features.Advances in neural information processing systems, 31, 2018

    Liudmila Prokhorenkova, Gleb Gusev, Aleksandr V orobev, Anna Veronika Dorogush, and Andrey Gulin. Catboost: unbiased boosting with categorical features.Advances in neural information processing systems, 31, 2018

    2018

  30. [38]

    Lightgbm: A highly efficient gradient boosting decision tree.Advances in neural information processing systems, 30, 2017

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. Lightgbm: A highly efficient gradient boosting decision tree.Advances in neural information processing systems, 30, 2017

    2017

  31. [39]

    Tabr: Unlocking the power of retrieval-augmented tabular deep learning.arXiv preprint arXiv:2307.14338, 2023

    Yury Gorishniy, Ivan Rubachev, Nikolay Kartashev, Daniil Shlenskii, Akim Kotelnikov, and Artem Babenko. Tabr: Unlocking the power of retrieval-augmented tabular deep learning.arXiv preprint arXiv:2307.14338, 2023

    2023 arXiv

  32. [40]

    Modern neighborhood components analysis: A deep tabular baseline two decades later.arXiv preprint arXiv:2407.03257, 2024

    Han-Jia Ye, Huai-Hong Yin, and De-Chuan Zhan. Modern neighborhood components analysis: A deep tabular baseline two decades later.arXiv preprint arXiv:2407.03257, 2024

    2024 arXiv

  33. [41]

    Saint: Improved neural networks for tabular data via row attention and contrastive pre-training

    Gowthami Somepalli, Micah Goldblum, Avi Schwarzschild, C Bayan Bruss, and Tom Goldstein. Saint: Improved neural networks for tabular data via row attention and contrastive pre-training. arXiv preprint arXiv:2106.01342, 2021

    2021 arXiv

  34. [42]

    Autogluon-tabular: Robust and accurate automl for structured data.arXiv preprint arXiv:2003.06505, 2020

    Nick Erickson, Jonas Mueller, Alexander Shirkov, Hang Zhang, Pedro Larroy, Mu Li, and Alexander Smola. Autogluon-tabular: Robust and accurate automl for structured data.arXiv preprint arXiv:2003.06505, 2020

    2003 arXiv

  35. [43]

    Scaling tree-based automated machine learning to biomedical big data with a feature set selector.Bioinformatics, 36(1):250–256, 2020

    Trang T Le, Weixuan Fu, and Jason H Moore. Scaling tree-based automated machine learning to biomedical big data with a feature set selector.Bioinformatics, 36(1):250–256, 2020

    2020

  36. [44]

    Algorithms for hyper- parameter optimization.Advances in neural information processing systems, 24, 2011

    James Bergstra, Rémi Bardenet, Yoshua Bengio, and Balázs Kégl. Algorithms for hyper- parameter optimization.Advances in neural information processing systems, 24, 2011

    2011

  37. [45]

    Optuna: A next-generation hyperparameter optimization framework

    Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. Optuna: A next-generation hyperparameter optimization framework. InProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2019

    2019

  38. [46]

    An inductive bias for tabular deep learning.Advances in Neural Information Processing Systems, 36, 2024

    Ege Beyazit, Jonathan Kozaczuk, Bo Li, Vanessa Wallace, and Bilal Fadlallah. An inductive bias for tabular deep learning.Advances in Neural Information Processing Systems, 36, 2024

    2024

  39. [47]

    Ex- celformer: A neural network surpassing gbdts on tabular data.arXiv preprint arXiv:2301.02819, 2023

    Jintai Chen, Jiahuan Yan, Qiyuan Chen, Danny Ziyi Chen, Jian Wu, and Jimeng Sun. Ex- celformer: A neural network surpassing gbdts on tabular data.arXiv preprint arXiv:2301.02819, 2023

    2023 arXiv

  40. [48]

    Better by default: Strong pre-tuned mlps and boosted trees on tabular data.Advances in Neural Information Processing Systems, 37:26577–26658, 2024

    David Holzmüller, Léo Grinsztajn, and Ingo Steinwart. Better by default: Strong pre-tuned mlps and boosted trees on tabular data.Advances in Neural Information Processing Systems, 37:26577–26658, 2024

    2024

  41. [49]

    Nonuniform fast fourier transforms using min-max interpolation.IEEE transactions on signal processing, 51(2):560–574, 2003

    Jeffrey A Fessler and Bradley P Sutton. Nonuniform fast fourier transforms using min-max interpolation.IEEE transactions on signal processing, 51(2):560–574, 2003

    2003

  42. [50]

    Accurate predictions on small data with a tabular foundation model.Nature, 637(8045):319–326, 2025

    Noah Hollmann, Samuel Müller, Lennart Purucker, Arjun Krishnakumar, Max Körfer, Shi Bin Hoo, Robin Tibor Schirrmeister, and Frank Hutter. Accurate predictions on small data with a tabular foundation model.Nature, 637(8045):319–326, 2025. 12 Appendix A Broader Societal Impact S...

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.