Pith. sign in

REVIEW 4 major objections 4 minor 38 references

Harnessing LLMs Explanations to Boost Surrogate Models in Tabular Data Classification

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Explanations generated once by GPT-3.5 can be turned into filtered demonstrations and rationale text that let a 7-billion-parameter surrogate model beat all baselines, including ChatGPT-driven in-context learning, on four tabular…

desk verdict The idea is sensible, but the method as written cannot produce the reported results — Eq.3 with n=5 and p=0.85 selects no features. read the letter →

arxiv 2505.05744 v1 pith:2R5NBF62 submitted 2025-05-09 cs.LG cs.CL

Ruxue Shi , Hengrui Gu , Xu Shen , Xin Wang This is my paper · ORCID
classification cs.LGcs.CL
keywords in-contextlearningtabulardataclassificationsurrogatelanguagemodelposthocexplanationsdemonstrationselectionfew-shotlargemodelsinterpretability
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 tries to establish that a small, locally deployable 7-billion-parameter language model can beat both classical tabular learners and direct few-shot prompting of a much larger API-only model, provided it is fed demonstrations selected and filtered by that larger model's post hoc explanations. The proposed framework generates feature-attribution explanations for 100 candidate rows, aggregates them into per-feature importance scores, keeps only the features that clear a threshold, and uses the cleaned rows to choose four demonstrations. Each selected demonstration has its explanation appended as a rationale, and the surrogate model is prompted to answer with a label and an explanation. Across the bank, creditg, heart, and income datasets, the paper reports 65.06% average 4-shot accuracy, 5.31 points above the strongest baseline, and higher accuracy than ChatGPT-based in-context learning while using 100 API calls instead of 2,400. The result matters because it suggests the expensive model can be consulted once, offline, while a cheap open model carries the repeated inference load.

What carries the argument

The load-bearing object is the post hoc explanation set $e_i$, a short list of feature names an LLM produces for candidate row $i$. From $M=100$ such sets the framework computes $g(w_j)=\frac{1}{nM}\sum_{i=1}^{M} \mathbf{1}\{w_j\in e_i\}$ (Eq.~3), the frequency with which feature $w_j$ is named; the top-$q$ features whose importance clears threshold $p=0.85$ form the allowed vocabulary $W$ (Eq.~4). Every candidate's serialized text is filtered down to sentences mentioning $W$ (Eq.~5), then embedded with SentenceBERT (Eq.~6). The demonstration score $s_i$ comes from one of four selection rules—K-Means distance, cosine similarity, Euclidean distance, or Manhattan distance (Eqs.~7--10)—and the top-$k$ rows are kept (Eq.~11). The mechanism does two jobs at once: filtering removes spurious features before embedding, so similarity is computed on a cleaner representation, and appending $e_i$ to each selected demonstration (Eq.~12) gives the SLM a rationale to imitate when it outputs a prediction plus explanation (Eqs.~13--14).

What would settle it

Replace GPT-3.5's explanations with uniformly random feature subsets of the same average size, keep the same p=0.85 filter and cosine selection, and rerun the four datasets; if average 4-shot accuracy stays near 65.06%, the explanation content is not carrying the gain, whereas a large drop would confirm it is.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the bottleneck in few-shot tabular classification with language models is not the size of the predictor but the information that reaches it. The authors build a three-stage loop: GPT-3.5 writes, for each of $M=100$ candidate rows, the $n=5$ feature names it judges most important for the answer; those explanations are aggregated into per-feature importance scores; features whose aggregate importance clears the threshold $p=0.85$ are kept and all other feature sentences are stripped from the serialized rows; and the cleaned rows are embedded and scored against the test row to choose the $k=4$ demonstrations. Those demonstrations are shown to Llama2-7B with their explanations appended as rationales, and the surrogate replies with both a class label and an explanation. The reported result is 65.06% average 4-shot accuracy on bank, creditg, heart, and income, 5.31 points above the strongest baseline, and higher accuracy than ChatGPT-based in-context learning while using 100 API calls instead of 2,400.

Load-bearing premise

The pipeline rests on the assumption that the feature names GPT-3.5 writes for 100 candidate rows form a reliable map of which features drive the label, so that deleting every feature absent from that map (threshold p=0.85) removes noise rather than signal; the paper does not measure the quality or seed-to-seed stability of those explanations.

Editorial extensions

If this is right

  • A 7B surrogate running this pipeline reaches 65.06% average 4-shot accuracy on bank, creditg, heart, and income, topping every baseline tested.
  • Each component matters: removing the appended explanations drops the average by about 8 points, removing the explanation-guided selection drops it by about 8 points, and removing both returns to the 53.94% plain in-context-learning baseline.
  • The framework moves almost all API cost to a one-time offline explanation stage: the Llama2-7B variant uses 100 GPT-3.5 calls, versus 2,400 for ChatGPT-based in-context learning, and still reports higher average accuracy (65.06% vs 59.90%).
  • Predictions arrive with an explanation string from the surrogate, so the interpretability is part of the output rather than a separate post-processing step.
  • Under cosine-similarity selection, the method beats random demonstration selection on all four datasets, with or without the feature filter.

Reading between the lines

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

  • [Editorial inference] The filter-and-select stage is effectively a feature-selection regularizer; comparing explanation-guided filtering against equal-sized random or variance-based feature deletion would show whether the explanatory content itself, rather than simple dimensionality reduction, drives the gain.
  • [Editorial inference] Since the expensive model is queried only 100 times up front, the same loop could plausibly be run with a small local explainer in place of GPT-3.5; if accuracy held, the method would need no commercial API after the warm-up stage.
  • [Editorial inference] The reported average covers four binary datasets, so the next natural test is whether the margin survives multi-class or high-cardinality categorical tables, where attribution quality is harder to judge.
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.

Referee Report

4 major / 4 minor

Summary. The paper introduces a three-stage in-context learning framework for tabular classification. In stage 1, GPT-3.5 generates post hoc feature-attribution explanations (n words) for M=100 candidate demonstration question-answer pairs. In stage 2, explanations are aggregated into feature importance scores g(w_j) (Eq. 3), a threshold p and top-q selection define a feature set W (Eq. 4), and serialized demonstrations are filtered to W (Eq. 5), embedded with SentenceBERT, and scored by cluster/cosine/Euclidean/Manhattan methods to select k-shot demonstrations. In stage 3, selected demonstrations are augmented with their explanations and fed to Llama2-7B, whose outputs are parsed into answer and explanation. Experiments on bank, creditg, heart, and income report 4-shot accuracy of 65.06% on average, a 5.31% improvement over the best average baseline, lower API usage than AO(ChatGPT), and ablations attributing gains to both explanation augmentation and guided selection.

Significance. The core idea is reasonable and of practical interest: use a strong LLM once to generate explanations that guide a locally deployable SLM, thereby reducing API calls while retaining interpretable rationales. The paper explicitly compares API usage and includes ablations and hyperparameter analysis. However, the current evidence is thin: four small-to-medium tabular datasets, one explanation generator (GPT-3.5), one surrogate (Llama2-7B), no statistical significance testing, and no code or data release. More importantly, the formal definition of the filtering step is inconsistent with the stated hyperparameters, so the reported numbers cannot be reproduced as written. If that inconsistency is resolved and the claims are re-validated, the framework would be a useful contribution; in its current form, the central empirical claim needs substantial revision.

major comments (4)
  1. [Section 3.3, Eqs. (3)-(5), and Section 4.2] Eq. (3) defines g(w_j) = (1/(n*M)) * sum_i 1{w_j in e_i}. Since each explanation e_i contains exactly n words, the maximum possible value of g(w_j) is 1/n. With n=5 and threshold p=0.85 (Section 4.2), no feature can satisfy the threshold, so W is empty. Eq. (5) then removes every feature sentence from S, leaving degenerate demonstrations; the selection and prediction stages of Eqs. (6)-(13) cannot yield the nondegenerate accuracies in Tables 3 and 5. Either g is normalized differently (e.g., by M), p is interpreted as a quantile, or an undocumented filtering criterion was used. The authors must correct the formal definition and provide code or a detailed algorithm trace showing a non-empty W for all datasets and seeds.
  2. [Section 4.4, Table 3] The sentence 'achieving a more than 5% performance improvement for all settings' is contradicted by the table. Per dataset, the improvements over the best baseline are 0.26 points on bank (55.26 vs SCARF 55.00), 0.00 on creditg (64.33 vs STUNT 64.33), 0.06 on heart (73.06 vs RF 73.00), and 0.58 on income (67.58 vs LR 67.00). The 5.31% figure refers only to the difference between the average accuracy and the average of the LR baseline. Moreover, several of these gaps are smaller than the reported standard deviations; no significance test is provided. The performance advantage claim must be restated and supported with error bars or statistical tests.
  3. [Section 4.5, Table 4] The text states that our approach 'achieves a higher average accuracy (6.18% improvement over AO(ChatGPT))', but Table 4 reports 65.06 vs 59.90, a difference of 5.16 percentage points. This numerical inconsistency must be corrected.
  4. [Section 3.2 and Section 4] The paper claims interpretability as a core contribution (Challenge 3 in the Introduction), yet no evaluation of explanation quality, faithfulness, or stability is reported. Since the whole framework depends on GPT-3.5 explanations being informative and stable across the M=100 candidate pool, the absence of any analysis of explanation correctness or seed sensitivity leaves a central assumption unverified. Please add quantitative or qualitative evaluation of explanations and an analysis of how W varies across candidate pools and seeds.
minor comments (4)
  1. [Eq. (4)] The notation Topq((w_i, g(w_i), p)) is ambiguous: it should specify whether p is an absolute cutoff on g, a rank quantile, or something else; this ambiguity is directly related to Major Comment 1.
  2. [Tables 3 and 5] Standard deviations are given as subscripts, but in the plain-text rendering they run into the values (e.g., '51.0013.00'), making the tables hard to read. Please format them with clear separators.
  3. [Figures 3 and 4] There are typos in the figure labels and captions: 'Eistance' should be 'Distance', 'credigt' should be 'creditg', and 'demonstrates' appears where 'demonstrations' is intended.
  4. [Table 4] Please clarify how the API usage counts are computed for each row, e.g., 100 explanation-generation calls for ours(Llama2-7B), 2400 inference calls for AO(ChatGPT), and 2500 for ours(ChatGPT).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the GPT-3.5 explanations are external inputs, and SLM accuracy is evaluated against ground-truth labels.

full rationale

The derivation chain is self-contained and not circular. Candidate demonstrations are randomly sampled from the training portion of external datasets (Bank, creditg, Heart, Income); GPT-3.5 generates post hoc explanations from those Q-A pairs (Eq. 2). The importance statistic g(w_j) (Eq. 3), the feature set W (Eq. 4), and the filter (Eq. 5) operate only on these candidate explanations and the serialized features; no test label or SLM output is fed back into the explanation generator. Demonstration selection (Eqs. 6-11) uses BERT embeddings and similarity/diversity scores, and the final prediction (Eq. 13) is produced by Llama2-7B conditioned on the k-shot prompt; accuracy is measured against external labels. Reusing the same explanations for both selection and as in-context rationales is a design choice, not a logical reduction: the explanations do not encode the test answer, and the SLM's answer is not defined in terms of them. There are no load-bearing self-citations and no imported uniqueness theorems. The literal reading of Eq. 3 with n=5 gives max g(w_j)=0.2, which would make the p=0.85 threshold in Section 4.2 empty; this is a formal normalization/consistency concern that could affect reproducibility, but it is not a circular derivation—the reported numbers are not forced by the paper's definitions, they are simply unexplained under that literal reading. The hyperparameter sweep in Figure 4 (p ranging to 1) suggests a frequency or cumulative interpretation was intended. The paper's central claim therefore does not reduce to its inputs.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The framework introduces no new theoretical entities. The main cost is empirical: a set of hand-chosen hyperparameters (n, p, q, k, M) and a strong reliance on the unexplained behavior of GPT-3.5 and Llama2-7B as black-box tools. The explanation sets e_i are the only new objects, and they are treated as ground truth without validation.

free parameters (5)
  • n = 5
    Number of post hoc explanation features requested from the LLM (Section 3.2, Section 4.8). Chosen by the authors; the heart dataset shows strong sensitivity to this value.
  • p = 0.85
    Filtering threshold for feature importance in Eq. 4 (Section 3.3). Chosen by the authors; Section 4.8 shows it has modest effect across datasets.
  • q = top-q features not explicitly set
    Number of top features selected in Eq. 4. The definition is ambiguous: W is defined via Topq with threshold p, but the exact q value is never given.
  • k = 4
    Number of few-shot demonstrations used in inference (Section 4.2, Eq. 11). Chosen by the authors.
  • M = 100
    Number of candidate demonstrations sampled randomly from training data (Section 3.1, Table 2). Chosen to limit API calls.
assumptions (3)
  • domain assumption GPT-3.5's feature-attribution explanations, averaged over 100 candidate demonstrations, identify features that are causally meaningful for the tabular prediction task.
    This is the load-bearing premise of the whole framework, entering at Eq. 2 and Eq. 3. The paper provides no evidence that the extracted words correspond to task-relevant features, only downstream accuracy.
  • domain assumption Filtering the serialized text to features in W and embedding with SentenceBERT produces representations whose similarity to the test sample selects useful demonstrations.
    Entering at Eq. 5, Eq. 6, and Eq. 8. This assumes that text embedding similarity after feature filtering is a good proxy for demonstration usefulness, which is an empirical claim tested only on four datasets.
  • domain assumption Llama2-7B can follow the 4-shot format and output a parsable label and explanation.
    Assumed at Eq. 13 and Eq. 14. The paper reports no parsing failure rate or output format validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Harnessing LLMs Explanations to Boost Surrogate Models in Tabular Data Classification." pith.science (2026). https://pith.science/paper/2R5NBF62

@misc{pith2026250505744,
  author       = {Pith},
  title        = {Pith review of: Harnessing LLMs Explanations to Boost Surrogate Models in Tabular Data Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2R5NBF62}},
  note         = {Machine review of arXiv:2505.05744}
}
read the original abstract

Large Language Models (LLMs) have shown remarkable ability in solving complex tasks, making them a promising tool for enhancing tabular learning. However, existing LLM-based methods suffer from high resource requirements, suboptimal demonstration selection, and limited interpretability, which largely hinder their prediction performance and application in the real world. To overcome these problems, we propose a novel in-context learning framework for tabular prediction. The core idea is to leverage the explanations generated by LLMs to guide a smaller, locally deployable Surrogate Language Model (SLM) to make interpretable tabular predictions. Specifically, our framework mainly involves three stages: (i) Post Hoc Explanation Generation, where LLMs are utilized to generate explanations for question-answer pairs in candidate demonstrations, providing insights into the reasoning behind the answer. (ii) Post Hoc Explanation-Guided Demonstrations Selection, which utilizes explanations generated by LLMs to guide the process of demonstration selection from candidate demonstrations. (iii) Post Hoc Explanation-Guided Interpretable SLM Prediction, which utilizes the demonstrations obtained in step (ii) as in-context and merges corresponding explanations as rationales to improve the performance of SLM and guide the model to generate interpretable outputs. Experimental results highlight the framework's effectiveness, with an average accuracy improvement of 5.31% across various tabular datasets in diverse domains.

Figures

Figures reproduced from arXiv: 2505.05744 by the authors.

Figure 1
Figure 1. Previous approaches to context learning with LLMs, as depicted on the top side of the figure, directly utilize tabular data to prompt the LLMs for predictions. This process necessitates frequent and extensive access to large language models (LLMs), resulting in substantial computational and efficiency demands. In contrast, our method depict at the bottom, presents a more efficient approach, which uses LLM generated … view at source ↗
Figure 2
Figure 2. Overview of our method. We begin by randomly selecting a small number of samples from the tabular dataset to form candidate demonstration sets. These samples are then converted into text and fed into an LLM, which generates post hoc explana￾tions. These explanations are used to guide the selection of relevant demonstrations and to facilitate the generation of interpretable predictions by the SLM. model. The LLMs the… view at source ↗
Figure 3
Figure 3. Comparison of different demonstration selection algorithms and baselines (ran￾dom selection) on four datasets. similarity algorithms (Euclidean distance and Manhattan distance), the perfor￾mance of the unfiltered(w/o filter) approach is occasionally inferior to random selection on certain datasets. However, with the integration of filtering(filter), all these algorithms outperform random selection consistently acros… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Effect of filter threshold p and the number of post hoc explanations n. The accuracy (acc) across all datasets is reported. strate relatively stable acc scores. Conversely, the heart dataset exhibits sub￾stantial variability, particularly a marked drop in acc at n = 6,…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 17 canonical work pages

  1. [1]

    In: Pro- ceedings of the AAAI conference on artificial intelligence

    Arik, S.Ö., Pfister, T.: Tabnet: Attentive interpretable tabular learning. In: Pro- ceedings of the AAAI conference on artificial intelligence. vol. 35, pp. 6679–6687 (2021)

    2021

  2. [2]

    Asuncion, A., Newman, D.: Uci machine learning repository (2007)

    work page 2007

  3. [3]

    arXiv preprint arXiv:2002.07971 (2020)

    Badirli, S., Liu, X., Xing, Z., Bhowmik, A., Doan, K., Keerthi, S.S.: Gradient boosting neural networks: Grownet. arXiv preprint arXiv:2002.07971 (2020)

    arXiv 2020

  4. [4]

    arXiv preprint arXiv:2106.15147 (2021)

    Bahri,D.,Jiang,H.,Tay,Y.,Metzler,D.:Scarf:Self-supervisedcontrastivelearning using random feature corruption. arXiv preprint arXiv:2106.15147 (2021)

    arXiv 2021

  5. [5]

    In: International Conference on Learning Representations (2022)

    Bahri, D., Jiang, H., Tay, Y., Metzler, D.: Scarf: Self-supervised contrastive learn- ing using random feature corruption. In: International Conference on Learning Representations (2022)

    work page 2022

  6. [6]

    Machine learning45, 5–32 (2001)

    Breiman, L.: Random forests. Machine learning45, 5–32 (2001)

    2001

  7. [7]

    arXiv preprint arXiv:2310.18541 (2023)

    Chen, S., Wu, J., Hovakimyan, N., Yao, H.: Recontab: Regularized contrastive representation learning for tabular data. arXiv preprint arXiv:2310.18541 (2023)

    arXiv 2023

  8. [8]

    In: Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining

    Chen, T., Guestrin, C.: Xgboost: A scalable tree boosting system. In: Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining. pp. 785–794 (2016)

    2016

Show all 38 references
  1. [9]

    Advances in Neural Information Processing Systems35, 11763–11784 (2022)

    Dinh, T., Zeng, Y., Zhang, R., Lin, Z., Gira, M., Rajput, S., Sohn, J.y., Pa- pailiopoulos, D., Lee, K.: Lift: Language-interfaced fine-tuning for non-language machine learning tasks. Advances in Neural Information Processing Systems35, 11763–11784 (2022)

    2022

  2. [10]

    In: 2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE)

    Fan, Z., Gao, X., Mirchev, M., Roychoudhury, A., Tan, S.H.: Automated repair of programs from large language models. In: 2023 IEEE/ACM 45th International Conference on Software Engineering (ICSE). pp. 1469–1481. IEEE (2023)

    2023

  3. [11]

    In: Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue

    Finch, S.E., Paek, E.S., Choi, J.D.: Leveraging large language models for auto- mated dialogue analysis. In: Proceedings of the 24th Meeting of the Special Interest Group on Discourse and Dialogue. pp. 202–215 (2023)

    2023

  4. [12]

    Advances in Neural Information Processing Systems34, 18932–18943 (2021)

    Gorishniy, Y., Rubachev, I., Khrulkov, V., Babenko, A.: Revisiting deep learning models for tabular data. Advances in Neural Information Processing Systems34, 18932–18943 (2021)

    2021

  5. [13]

    In: Statistical models in S, pp

    Hastie, T.J., Pregibon, D.: Generalized linear models. In: Statistical models in S, pp. 195–247. Routledge (2017)

    2017

  6. [14]

    In: Proceedings of 3rd international conference on document analysis and recognition

    Ho, T.K.: Random decision forests. In: Proceedings of 3rd international conference on document analysis and recognition. vol. 1, pp. 278–282. IEEE (1995)

    1995

  7. [15]

    In: The Eleventh International Conference on Learning Representations (2023)

    Hollmann, N., Müller, S., Eggensperger, K., Hutter, F.: Tabpfn: A transformer that solves small tabular classification problems in a second. In: The Eleventh International Conference on Learning Representations (2023)

    2023

  8. [16]

    arXiv preprint arXiv:2012.06678 (2020)

    Huang, X., Khetan, A., Cvitkovic, M., Karnin, Z.: Tabtransformer: Tabular data modeling using contextual embeddings. arXiv preprint arXiv:2012.06678 (2020)

    2020 arXiv

  9. [17]

    Advances in neural information processing systems34, 23928– 23941 (2021)

    Kadra, A., Lindauer, M., Hutter, F., Grabocka, J.: Well-tuned simple nets excel on tabular datasets. Advances in neural information processing systems34, 23928– 23941 (2021)

    2021

  10. [18]

    Advances in neural information processing systems30 (2017)

    Ke, G., Meng, Q., Finley, T., Wang, T., Chen, W., Ma, W., Ye, Q., Liu, T.Y.: Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems30 (2017)

    2017

  11. [19]

    In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining

    Ke, G., Xu, Z., Zhang, J., Bian, J., Liu, T.Y.: Deepgbm: A deep learning framework distilled by gbdt for online prediction tasks. In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. pp. 384–394 (2019) R.Shi et al

    2019

  12. [20]

    Ke, G., Zhang, J., Xu, Z., Bian, J., Liu, T.Y.: Tabnn: A universal neural network solution for tabular data (2018)

    2018

  13. [21]

    Advances in neural information processing systems30 (2017)

    Klambauer, G., Unterthiner, T., Mayr, A., Hochreiter, S.: Self-normalizing neural networks. Advances in neural information processing systems30 (2017)

    2017

  14. [22]

    Circulation117(18), 2395–2399 (2008)

    LaValley, M.P.: Logistic regression. Circulation117(18), 2395–2399 (2008)

    2008

  15. [23]

    ACM Computing Surveys55(9), 1–35 (2023)

    Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., Neubig, G.: Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys55(9), 1–35 (2023)

    2023

  16. [24]

    Loh,W.Y.:Classificationandregressiontrees.Wileyinterdisciplinaryreviews:data mining and knowledge discovery1(1), 14–23 (2011)

    2011

  17. [25]

    Decision Support Systems62, 22–31 (2014)

    Moro, S., Cortez, P., Rita, P.: A data-driven approach to predict the success of bank telemarketing. Decision Support Systems62, 22–31 (2014)

    2014

  18. [26]

    In: Work- shop on Efficient Systems for Foundation Models@ ICML2023 (2023)

    Nam, J., Song, W., Park, S.H., Tack, J., Yun, S., Kim, J., Shin, J.: Semi-supervised tabular classification via in-context learning of large language models. In: Work- shop on Efficient Systems for Foundation Models@ ICML2023 (2023)

    2023

  19. [27]

    In: The Eleventh International Conference on Learning Representations (2023)

    Nam, J., Tack, J., Lee, K., Lee, H., Shin, J.: Stunt: Few-shot tabular learning with self-generated tasks from unlabeled tables. In: The Eleventh International Conference on Learning Representations (2023)

    2023

  20. [28]

    arXiv preprint arXiv:1909.06312 (2019)

    Popov, S., Morozov, S., Babenko, A.: Neural oblivious decision ensembles for deep learning on tabular data. arXiv preprint arXiv:1909.06312 (2019)

    2019 arXiv

  21. [29]

    Advances in neural information pro- cessing systems 31 (2018)

    Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A.V., Gulin, A.: Catboost: unbiased boosting with categorical features. Advances in neural information pro- cessing systems 31 (2018)

    2018

  22. [30]

    arXiv preprint arXiv:2304.13188 (2023)

    Slack, D., Singh, S.: Tablet: Learning from instructions for tabular data. arXiv preprint arXiv:2304.13188 (2023)

    2023 arXiv

  23. [31]

    arXiv preprint arXiv:2106.01342 (2021)

    Somepalli, G., Goldblum, M., Schwarzschild, A., Bruss, C.B., Goldstein, T.: Saint: Improved neural networks for tabular data via row attention and contrastive pre- training. arXiv preprint arXiv:2106.01342 (2021)

    2021 arXiv

  24. [32]

    In: Pro- ceedings of the 28th ACM international conference on information and knowledge management

    Song, W., Shi, C., Xiao, Z., Duan, Z., Xu, Y., Zhang, M., Tang, J.: Autoint: Automatic feature interaction learning via self-attentive neural networks. In: Pro- ceedings of the 28th ACM international conference on information and knowledge management. pp. 1161–1170 (2019)

    2019

  25. [33]

    arXiv preprint arXiv:2302.13971 (2023)

    Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al.: Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023)

    2023 arXiv

  26. [34]

    In: Proceedings of the web conference 2021

    Wang, R., Shivanna, R., Cheng, D., Jain, S., Lin, D., Hong, L., Chi, E.: Dcn v2: Improved deep & cross network and practical lessons for web-scale learning to rank systems. In: Proceedings of the web conference 2021. pp. 1785–1797 (2021)

    2021

  27. [35]

    arXiv preprint arXiv:2206.07682 (2022)

    Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D., Metzler, D., et al.: Emergent abilities of large language models. arXiv preprint arXiv:2206.07682 (2022)

    2022 arXiv

  28. [36]

    Advances in Neural Information Processing Systems35, 24824–24837 (2022)

    Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q.V., Zhou, D., et al.: Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems35, 24824–24837 (2022)

    2022

  29. [37]

    Advances in Neural Information Processing Systems33, 11033–11043 (2020)

    Yoon, J., Zhang, Y., Jordon, J., van der Schaar, M.: Vime: Extending the suc- cess of self-and semi-supervised learning to tabular domain. Advances in Neural Information Processing Systems33, 11033–11043 (2020)

    2020

  30. [38]

    In: The Eleventh International Conference on Learning Representations

    Zhang, Z., Zhang, A., Li, M., Smola, A.: Automatic chain of thought prompting in large language models. In: The Eleventh International Conference on Learning Representations

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.