Pith. sign in

REVIEW 3 major objections 6 minor 41 references

Finding Needles in Emb(a)dding Haystacks: Legal Document Retrieval via Bagging and SVR Ensembles

T0 review · 3 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A bagged ensemble of 35 Support Vector Regressors over Longformer embeddings retrieves German legal passages at recall 0.849, beating published baselines without fine-tuning any deep model.

desk verdict The headline recall number is per-candidate classification recall on a 50-neighbor subset, not retrieval recall; the paper's central comparison is apples-to-oranges, but the method is a reasonable cheap baseline worth a re-run. read the letter →

arxiv 2501.05018 v1 pith:4J2R3ZOY submitted 2025-01-09 cs.IR cs.AI

Kevin Bönisch , Alexander Mehler This is my paper · ORCID
classification cs.IRcs.AI
keywords legalinformationretrievalsupportvectorregressionwordembeddingsbaggingensembleGerDaLIRLongformerdensepassagek-nearestneighbors
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 claims that German legal passage retrieval can be improved by treating it as a set of binary needle-in-a-haystack subtasks, each learned by a small Support Vector Regression model over transformer embeddings. On the GerDaLIR benchmark it reports recall of 0.849, above the best published baseline values of 0.803 and 0.829, without fine-tuning any deep learning model. The result matters because legal retrieval in German has few dedicated tools, and a transparent classical-machine-learning route that reaches neural-level recall would give practitioners an explainable alternative. The same architecture could be upgraded by swapping in better embedding encoders or enlarging the candidate neighborhood.

What carries the argument

The load-bearing object is the bagged SVR ensemble: 35 Support Vector Regressors with RBF kernels, each trained on an overlapping subset of the collection's Longformer embedding space, with predictions combined by voting. Features are formed by concatenating the query's 768-dimensional embedding with each candidate passage's embedding, producing a 1,536-dimensional vector for each of the $k=50$ nearest neighbors. The SVR is trained as a binary regressor whose target is 1 for the gold passage and 0 for the $k-1$ negatives, an instance of the needle-in-a-haystack formulation. The k-NN candidate step determines what the ensemble can ever see, so the whole pipeline's ceiling is fixed before the regressors are trained.

What would settle it

Measure the top-50 candidate recall of Longformer alone on the GerDaLIR test split: count how often the gold passage appears among the 50 nearest neighbors of its query. If this number is close to 0.849 or below it, the reported recall is set by the candidate step, not by the SVR ensemble, and changing $k$ would change the headline comparison.

Watch

Extended reading notes

Core claim

The central claim is that a voting ensemble of 35 bagged SVR models, each trained on a 60%-overlapping subset of the Longformer embedding space, retrieves relevant German legal passages at recall 0.849, beating the published TF-IDF, BM25, and BERT/ELECTRA re-ranking baselines on GerDaLIR. For every query the method takes its 50 nearest passage neighbors, concatenates query and passage embeddings into a 1,536-dimensional feature vector, and trains each SVR to separate the single relevant passage from the 49 distractors. No deep model is trained or fine-tuned; the only learned component is the classical SVR ensemble, which the paper argues is more transparent than a fine-tuned transformer for legal search.

Load-bearing premise

The method assumes that for every query, the truly relevant passage is among the 50 closest passages in the Longformer embedding space; if that fails, no amount of SVR training can retrieve the passage.

Editorial extensions

If this is right

  • If the reported recall is accepted, a classical SVR ensemble can outperform fine-tuned neural re-rankers on German legal retrieval, making strong retrieval achievable without GPU-heavy fine-tuning of large models.
  • The bottleneck moves to first-stage candidate recall: with $k=50$, any relevant passage outside the nearest-neighbor set is unreachable, so improving recall requires enlarging $k$ or better embeddings rather than better regressors.
  • Because the method treats retrieval as many independent binary subtasks, it can be parallelized and reproduced with standard machine learning tooling, and each per-subset model can be inspected individually.
  • The same pipeline is directly portable to other languages and document collections by replacing the embedding encoder, since no part of the training depends on German-specific resources.

Reading between the lines

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

  • Editorial inference: the paper's own exploratory observation that embedding distance is dominated by text length suggests a cheap testable extension — chunk queries and passages to equal length before embedding — which could raise candidate recall without changing the ensemble.
  • Editorial inference: because only Recall is reported, it is an open question how the ensemble ranks at the top of the list; measuring MRR@10 and nDCG@20 the way the baselines were measured would show whether the recall gain comes with usable ranking quality.
  • Editorial inference: comparing a single embedding model's top-50 candidate recall against the 0.849 ensemble recall would isolate how much of the gain is due to SVR ranking versus Longformer embedding geometry; if candidate recall is already near 0.849, the SVRs add little.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes a lightweight legal document retrieval method for the German GerDaLIR dataset. The method first embeds all collection passages with a pre-trained Longformer model, then for each query selects the k=50 nearest neighbors in the embedding space. A bagged ensemble of Support Vector Regression (SVR) models, each trained on an overlapping subset of the collection, is then used to identify the relevant passage among these k candidates. The authors report a recall of 0.849 and claim this surpasses the GerDaLIR baselines (TF-IDF, BM25, and BERT-based re-ranking) whose recall values are 0.803 and 0.829, all without fine-tuning any deep learning models.

Significance. If the claimed improvement were established under a comparable evaluation protocol, the method would be a useful, transparent, and computationally light baseline for German legal information retrieval. The paper is written with a clear structure, and the release of source code is a positive step. However, the significance is currently undermined by a fundamental evaluation mismatch: the reported 0.849 recall appears to be a per-candidate classification recall on a skewed test set, not a query-level retrieval recall over the full collection, and the method's end-to-end performance is bounded by an unmeasured k=50 nearest-neighbor oracle. Until these issues are resolved, the central claim of improved retrieval recall is not supported.

major comments (3)
  1. [§3.3, §3.2, Table 1] The reported recall of 0.849 is not protocol-comparable to the baselines. The baseline metrics are Recall@100 and Recall@1000, i.e., the fraction of test queries for which a relevant passage appears in the top 100 or top 1000 retrieval results over the full collection. In contrast, the SVR ensemble's recall (Figure 4) is computed on the held-out split of the candidate-pair classification task, where positive examples are defined as the labeled relevant passage among the k=50 nearest Longformer neighbors. These are different quantities: the former is an end-to-end retrieval measure, the latter is a binary classification recall on a pre-filtered candidate set. The paper must report query-level retrieval recall over the full collection using the same cutoff-based protocol as the baselines, along with MRR and nDCG, or it cannot claim an improvement over the baselines.
  2. [§3.2, §4, Table 1 note] The method's retrieval ceiling is the fraction of test queries whose relevant passage lies among the k=50 nearest Longformer neighbors, a quantity the paper never reports. The text in Section 3.2 explicitly states, 'Our training assumes that the relevant passage to be retrieved is somewhere among these k passages,' yet no experiment verifies this assumption. Section 4 acknowledges that outlier passages are 'positioned far from their corresponding queries,' so the oracle recall could be substantially below the baseline recall values of 0.803 and 0.829. The Table 1 note that the baselines' Recall@100/1000 is 'a restriction that we do not impose' is misleading: the proposed method imposes a far stricter implicit restriction of 50 candidates. The authors should report the kNN oracle recall and, if it is below baseline values, explicitly state that the method is not a full retrieval system.
  3. [§3.2, §3.3, Figure 4] The evaluation is reported as a per-pair classification result, not a per-query retrieval result. The class-1 recall of 0.8527 in Figure 4, which is rounded to 0.849 in Table 1, reflects the fraction of positive candidate pairs correctly classified among the 50 neighbors, not the fraction of queries for which the method returns a relevant passage. The paper does not describe how the ensemble's per-model predictions are aggregated when more than one model produces a positive prediction, nor how the method handles queries with multiple relevant passages (the dataset labels 'at least one document and hence multiple passages' per query). Without a clear query-level aggregation rule and a query-level recall computation, the reported 0.849 cannot be interpreted as a retrieval recall.
minor comments (6)
  1. [Table 1] The row for the SVR ensemble ('Method Mode Recall') breaks the table's column structure and would be clearer if it reported the same metrics as the baselines (MRR@10, nDCG@20, Recall@100, Recall@1000) or explicitly stated that those metrics are not applicable.
  2. [§1] The phrase 'we publish our soure code' contains a typo; it should be 'source code.'
  3. [§2.2] The section title 'Supported Vector Machine Regression' is nonstandard; the usual term is 'Support Vector Regression.'
  4. [§3.1] The t-SNE plots are descriptive and useful, but the claim that 'the longformer_base embedding space seems to capture the context best' would be better supported by a quantitative comparison, such as average distance from query to relevant passage or recall at various k, rather than visual inspection alone.
  5. [§3.2] The bagging configuration (35 subsets, 60% overlap, k=50) and the SVR hyperparameters (C, epsilon, gamma) are stated as fixed choices without justification or sensitivity analysis; reporting these details and at least a small ablation would strengthen reproducibility.
  6. [§3.2] The paper does not specify the random seed, the exact split of the training/validation data, or whether the test split from GerDaLIR (mentioned in the text) is used for the final evaluation; a more precise description of the data partitioning is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the SVR ensemble recall is measured on a held-out split from labels; the protocol mismatch with GerDaLIR baselines is a validity issue, not a circular derivation.

full rationale

The paper's derivation chain is self-contained. The feature matrix is built by embedding each query and its k=50 nearest passages, labeling one passage positive per query and k-1 negative, and the SVR models are trained on these labels and evaluated on a 0.9/0.1 split (Section 3.2). The reported 0.849 recall is the class-1 recall from Figures 3-4, i.e., the fraction of true positive candidate passages that the ensemble flags; it is not defined in terms of the baseline Recall@100/Recall@1000 values and is not fitted to those numbers. The model can fail on this task (class-1 recall is 0.85 rather than 1.00), so the measurement has independent content. No load-bearing self-citation appears: the Longformer encoder and GerDaLIR are external prior work, and the paper's own exploratory analysis is used only to choose the encoder, not to define the metric. The main caveat - that retrieval is capped by whether the relevant passage falls among the k=50 nearest neighbors, a bound the paper explicitly acknowledges in Section 3.2 and Section 4 - is an evaluation-protocol issue (the 0.849 recall is not query-level retrieval recall and is not directly comparable to the baselines), not a circularity: the assumption is stated explicitly and is not the source of the 0.849 value. I therefore assign score 0.

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

The central claim rests on several hand-chosen hyperparameters (k, subset count, overlap, encoder) and on unproven transfer assumptions about embeddings. No new theoretical entities are introduced; the method is empirical and its reported recall depends materially on these choices.

free parameters (5)
  • k (nearest-neighbor candidates) = 50
    Hand-chosen; sets the recall ceiling because passages outside the top 50 cannot be retrieved (Section 3.2).
  • s (number of bagging subsets) = 35
    Hand-chosen for this baseline; changes ensemble size and training cost (Section 3.2).
  • subset overlap fraction = 0.60
    Hand-chosen; controls redundancy across bagging subsets (Section 3.2).
  • SVR hyperparameters (C, epsilon, gamma) = not reported
    Not specified in the text; directly affect regression output and the reported recall (Sections 3.2-3.3).
  • encoder model choice = longformer_base
    Selected after exploratory t-SNE analysis on a 10,000-passage sample; a different encoder changes the whole embedding space (Section 3.1).
assumptions (5)
  • ad hoc to paper The relevant passage for each query lies within the k=50 nearest neighbors in the Longformer embedding space.
    Section 3.2 states: 'Our training assumes that the relevant passage to be retrieved is somewhere among these k passages.'
  • domain assumption Embedding-space proximity between query and passage is a valid relevance signal.
    Sections 2.1 and 3.1 motivate this with t-SNE plots, but provide no quantitative validation.
  • domain assumption An English-pretrained Longformer encoder transfers well enough to German legal passages.
    Section 4 acknowledges the encoder models 'primarily process English text'; the whole pipeline depends on this transfer.
  • domain assumption Binary SVR classification with one positive and k-1 negatives per query generalizes to the GerDaLIR test queries.
    Section 3.2 defines the training task; no theoretical or empirical evidence beyond the single test split is given.
  • domain assumption GerDaLIR labels and its separate test split are correctly used without leakage.
    The paper relies on the dataset's provided split (Sections 2.4 and 3.2) but does not audit for leakage.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Finding Needles in Emb(a)dding Haystacks: Legal Document Retrieval via Bagging and SVR Ensembles." pith.science (2026). https://pith.science/paper/4J2R3ZOY

@misc{pith2026250105018,
  author       = {Pith},
  title        = {Pith review of: Finding Needles in Emb(a)dding Haystacks: Legal Document Retrieval via Bagging and SVR Ensembles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4J2R3ZOY}},
  note         = {Machine review of arXiv:2501.05018}
}
read the original abstract

We introduce a retrieval approach leveraging Support Vector Regression (SVR) ensembles, bootstrap aggregation (bagging), and embedding spaces on the German Dataset for Legal Information Retrieval (GerDaLIR). By conceptualizing the retrieval task in terms of multiple binary needle-in-a-haystack subtasks, we show improved recall over the baselines (0.849 > 0.803 | 0.829) using our voting ensemble, suggesting promising initial results, without training or fine-tuning any deep learning models. Our approach holds potential for further enhancement, particularly through refining the encoding models and optimizing hyperparameters.

Figures

Figures reproduced from arXiv: 2501.05018 by the authors.

Figure 2
Figure 2. Modeling the flow of the needle-in-a-haystack training, we begin by partitioning the document space into several subsets (bagging), with each subset being assigned a separate SVR model for training. For each query in a subset, we identify the top 𝑘 nearest passages through their vector spaces and concatenate their embeddings into a single feature embedding. Consequently, each query is associated with 𝑘 − 1 negative … view at source ↗
Figure 3
Figure 3. Model Accuracy Metric Value Test set accuracy 0.9966 Precision 0.9987 Recall 0.8527 F1-score 0.9199 [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 18 canonical work pages

  1. [1]

    Salton, M

    G. Salton, M. McGill, Introduction to modern information retrieval, 1983. URL: https: //api.semanticscholar.org/CorpusID:43685115

    work page 1983

  2. [2]

    Sansone, G

    C. Sansone, G. Sperlí, Legal information retrieval systems: State-of-the-art and open issues, Information Systems 106 (2022) 101967

    work page 2022

  3. [3]

    Salton, Automatic Text Processing: The Transformation, Analysis, and Retrieval of Information by Computer, Addison Wesley, Reading, Massachusetts, 1989

    G. Salton, Automatic Text Processing: The Transformation, Analysis, and Retrieval of Information by Computer, Addison Wesley, Reading, Massachusetts, 1989

    work page 1989

  4. [4]

    Wehnert, V

    S. Wehnert, V. Sudhi, S. Dureja, L. Kutty, S. Shahania, E. W. De Luca, Legal norm retrieval with variations of the bert model combined with tf-idf vectorization, in: Proceedings of the Eighteenth International Conference on Artificial Intelligence and Law, ICAIL ’21, Association for Computing Machinery, New York, NY, USA, 2021, p. 285–294. URL: https://do...

    arXiv 2021

  5. [5]

    C. D. Manning, P. Raghavan, H. Schütze, Introduction to Information Retrieval, Cambridge University Press, 2008

    work page 2008

  6. [6]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, I. Polosukhin, Attention is all you need, 2023. URL: https://arxiv.org/abs/1706.03762. arXiv:1706.03762

    arXiv 2023

  7. [7]

    Nogueira, K

    R. Nogueira, K. Cho, Passage re-ranking with bert, 2020. URL: https://arxiv.org/abs/1901. 04085. arXiv:1901.04085

    arXiv 2020

  8. [8]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, Bert: Pre-training of deep bidirectional transformers for language understanding, 2019. URL: https://arxiv.org/abs/1810.04805. arXiv:1810.04805

    arXiv 2019

Show all 41 references
  1. [9]

    Y. Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. Zettlemoyer, V. Stoyanov, Roberta: A robustly optimized bert pretraining approach, 2019. URL: https: //arxiv.org/abs/1907.11692. arXiv:1907.11692

    2019 arXiv

  2. [10]

    P. He, X. Liu, J. Gao, W. Chen, Deberta: Decoding-enhanced bert with disentangled attention, 2021. URL: https://arxiv.org/abs/2006.03654. arXiv:2006.03654

    2021 arXiv

  3. [11]

    Chalkidis, M

    I. Chalkidis, M. Fergadiotis, P. Malakasiotis, N. Aletras, I. Androutsopoulos, Legal-bert: The muppets straight out of law school, 2020. URL: https://arxiv.org/abs/2010.02559. arXiv:2010.02559

    2020 arXiv

  4. [12]

    S. Yue, W. Chen, S. Wang, B. Li, C. Shen, S. Liu, Y. Zhou, Y. Xiao, S. Yun, X. Huang, Z. Wei, Disc-lawllm: Fine-tuning large language models for intelligent legal services, 2023. URL: https://arxiv.org/abs/2309.11325. arXiv:2309.11325

    2023 arXiv

  5. [13]

    T. B. Brown, B. Mann, N. Ryder, M. Subbiah, J. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell, S. Agarwal, A. Herbert-Voss, G. Krueger, T. Henighan, R. Child, A. Ramesh, D. M. Ziegler, J. Wu, C. Winter, C. Hesse, M. Chen, E. Sigler, M. Litwin, S. Gray, B. ...

    2020 arXiv

  6. [14]

    H. Luo, L. Specia, From understanding to utilization: A survey on explainability for large language models, 2024. URL: https://arxiv.org/abs/2401.12874. arXiv:2401.12874

    2024 arXiv

  7. [15]

    H. Zhao, H. Chen, F. Yang, N. Liu, H. Deng, H. Cai, S. Wang, D. Yin, M. Du, Explain- ability for large language models: A survey, 2023. URL: https://arxiv.org/abs/2309.01029. arXiv:2309.01029

    2023 arXiv

  8. [16]

    Rabelo, M.-Y

    J. Rabelo, M.-Y. Kim, R. Goebel, M. Yoshioka, Y. Kano, K. Satoh, Coliee 2020: Methods for legal document retrieval and entailment, in: N. Okazaki, K. Yada, K. Satoh, K. Mineshima (Eds.), New Frontiers in Artificial Intelligence, Springer International Publishing, Cham, 2021, p...

    2020

  9. [17]

    Sugathadasa, B

    K. Sugathadasa, B. Ayesha, N. de Silva, A. S. Perera, V. Jayawardana, D. Lakmal, M. Perera, Legal document retrieval using document vector embeddings and deep learning, 2018. URL: https://arxiv.org/abs/1805.10685. arXiv:1805.10685

    2018 arXiv

  10. [18]

    Mandal, K

    A. Mandal, K. Ghosh, A. Bhattacharya, A. Pal, S. Ghosh, Overview of the fire 2017 irled track: Information retrieval from legal documents, in: Fire, 2017. URL: https://api. semanticscholar.org/CorpusID:39265594

    2017

  11. [19]

    Wrzalik, D

    M. Wrzalik, D. Krechel, GerDaLIR: A German dataset for legal information retrieval, in: N. Aletras, I. Androutsopoulos, L. Barrett, C. Goanta, D. Preotiuc-Pietro (Eds.), Proceedings of the Natural Legal Language Processing Workshop 2021, Association for Computational Linguisti...

    2021 doi

  12. [20]

    Hearst, S

    M. Hearst, S. Dumais, E. Osuna, J. Platt, B. Scholkopf, Support vector machines, IEEE Intelligent Systems and their Applications 13 (1998) 18–28. doi:10.1109/5254.708428

    1998

  13. [21]

    Breiman, Bagging Predictors, Machine Learning 24 (1996) 123–140

    L. Breiman, Bagging Predictors, Machine Learning 24 (1996) 123–140. doi: 10.1007/ BF00058655

    1996

  14. [22]

    Breiman, Random forests, Machine Learning 45 (2001) 5–32

    L. Breiman, Random forests, Machine Learning 45 (2001) 5–32. URL: https://doi.org/10. 1023/A:1010933404324. doi:10.1023/A:1010933404324

    2001 doi

  15. [23]

    Salton, C

    G. Salton, C. Buckley, Term weighting approaches in automatic text retrieval, Information Processing Management 24 (1988) 513–523

    1988

  16. [24]

    J. Beel, B. Gipp, S. Langer, C. Breitinger, Research-paper recommender systems: a literature survey, International Journal on Digital Libraries 17 (2016) 305–338. URL: https://doi.org/ 10.1007/s00799-015-0156-0. doi: 10.1007/s00799-015-0156-0

    2016 doi

  17. [25]

    Nguyen, H.-Y

    H.-T. Nguyen, H.-Y. T. Vuong, P. M. Nguyen, B. T. Dang, Q. M. Bui, S. T. Vu, C. M. Nguyen, V. Tran, K. Satoh, M. L. Nguyen, Jnlp team: Deep learning for legal processing in coliee 2020, 2020. URL: https://arxiv.org/abs/2011.08071. arXiv:2011.08071

    2020 arXiv

  18. [26]

    M.-Y. Kim, J. Rabelo, R. Goebel, M. Yoshioka, Y. Kano, K. Satoh, Coliee 2022 summary: Methods for legal document retrieval and entailment, in: Y. Takama, K. Yada, K. Satoh, S. Arai (Eds.), New Frontiers in Artificial Intelligence, Springer Nature Switzerland, Cham, 2023, pp. 51–67

    2022

  19. [27]

    Kamalloo, X

    E. Kamalloo, X. Zhang, O. Ogundepo, N. Thakur, D. Alfonso-Hermelo, M. Rezagholizadeh, J. Lin, Evaluating embedding apis for information retrieval, 2023. URL: https://arxiv.org/ abs/2305.06300. arXiv:2305.06300

    2023 arXiv

  20. [28]

    Galke, A

    L. Galke, A. Saleh, A. Scherp, Word embeddings for practical information retrieval, INFOR- MATIK 2017, 2017. doi:10.18420/in2017_215

    2017 doi

  21. [29]

    Salton, A

    G. Salton, A. Wong, C. S. Yang, A vector space model for automatic indexing, Com- mun. ACM 18 (1975) 613–620. URL: https://doi.org/10.1145/361219.361220. doi:10.1145/ 361219.361220

    1975

  22. [30]

    Mikolov, K

    T. Mikolov, K. Chen, G. Corrado, J. Dean, Efficient estimation of word representations in vector space, 2013. URL: https://arxiv.org/abs/1301.3781. arXiv:1301.3781

    2013 arXiv

  23. [31]

    V. N. Vapnik, The Nature of Statistical Learning Theory, Springer, New York, 1995

    1995

  24. [32]

    Drucker, C

    H. Drucker, C. J. C. Burges, L. Kaufman, A. Smola, V. Vapnik, Support vector regression machines, in: M. Mozer, M. Jordan, T. Petsche (Eds.), Advances in Neural Information Processing Systems, volume 9, MIT Press, 1996. URL: https://proceedings.neurips.cc/paper_ files/paper/19...

    1996

  25. [33]

    Sharafati, S

    A. Sharafati, S. B. Haji Seyed Asadollah, N. Al-Ansari, Application of bagging en- semble model for predicting compressive strength of hollow concrete masonry prism, Ain Shams Engineering Journal 12 (2021) 3521–3530. URL: https://www.sciencedirect. com/science/article/pii/S209...

    2021 doi

  26. [34]

    H. Qian, D. Ying, H. Lam, W. Yin, Bagging improves generalization exponentially, 2024. URL: https://arxiv.org/abs/2405.14741. arXiv:2405.14741

    2024

  27. [35]

    Van der Maaten, G

    L. Van der Maaten, G. Hinton, Visualizing data using t-sne., Journal of machine learning research 9 (2008)

    2008

  28. [36]

    Beltagy, M

    I. Beltagy, M. E. Peters, A. Cohan, Longformer: The long-document transformer, 2020. URL: https://arxiv.org/abs/2004.05150. arXiv:2004.05150

    2020 arXiv

  29. [37]

    Karpukhin, B

    V. Karpukhin, B. Oguz, S. Min, P. Lewis, L. Wu, S. Edunov, D. Chen, W.-t. Yih, Dense passage retrieval for open-domain question answering, in: B. Webber, T. Cohn, Y. He, Y. Liu (Eds.), Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNL...

    2020 doi

  30. [38]

    Y. Zhu, H. Yuan, S. Wang, J. Liu, W. Liu, C. Deng, H. Chen, Z. Dou, J.-R. Wen, Large language models for information retrieval: A survey, 2024. URL: https://arxiv.org/abs/2308.07107. arXiv:2308.07107

    2024

  31. [39]

    Tsuda, M

    T. Tsuda, M. Imaizumi, Benign overfitting of non-sparse high-dimensional linear regression with correlated noise, 2023. URL: https://arxiv.org/abs/2304.04037. arXiv:2304.04037

    2023 arXiv

  32. [40]

    A. Y. Ng, Feature selection, l1 vs. l2 regularization, and rotational invariance, in: Pro- ceedings of the Twenty-First International Conference on Machine Learning, ICML ’04, Association for Computing Machinery, New York, NY, USA, 2004, p. 78. URL: https://doi.org/10.1145/101...

    2004

  33. [41]

    Ghosh, A

    S. Ghosh, A. Dasgupta, A. Swetapadma, A study on support vector machine based linear and non-linear pattern classification, in: 2019 International Conference on Intelligent Sustainable Systems (ICISS), 2019, pp. 24–28. doi:10.1109/ISS1.2019.8908018

    2019

Pith tools

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