Class Distillation with Mahalanobis Contrast: An Efficient Training Paradigm for Pragmatic Language Understanding Tasks

Chenlu Wang; Ritwik Banerjee; Weimin Lyu

REVIEW 4 major objections 6 minor 66 references

Class Distillation with Mahalanobis Contrast: An Efficient Training Paradigm for Pragmatic Language Understanding Tasks

T0 review · 4 major / 6 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 ClaD, a one-epoch geometric training scheme, matches or beats LLMs on sarcasm, metaphor, and sexism detection.

desk verdict A genuinely new training/inference combo for pragmatic-language minority-class detection, but the efficiency claim needs seed-level variance before I'd trust the headline numbers. read the letter →

arxiv 2505.11829 v1 pith:YCM6XI5S submitted 2025-05-17 cs.CL

Chenlu Wang , Weimin Lyu , Ritwik Banerjee This is my paper · ORCID

classification cs.CL MSC 68T5062H30

keywords classdistillationMahalanobisdistancepragmaticlanguageunderstandingsarcasmdetectionmetaphorsexismBetadistributionlow-resourcelearning

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes a training paradigm, ClaD, for the common but hard setup of detecting a small target class—sarcastic, metaphorical, or sexist language—inside a large, diverse 'everything else' background. The claim is that the target class has a compact, near-normal geometric manifold, and that exploiting this with a Mahalanobis-distance contrast loss plus a Beta-distribution decision rule lets a small encoder match or beat large language models after a single epoch of fine-tuning. On three benchmarks, ClaD outperforms standard fine-tuned encoders and, with orders of magnitude fewer parameters, matches or exceeds several LLMs in few-shot, low-resource, and identical-training-data comparisons. The paper argues this shows that geometric modeling of the target class can substitute for scale in pragmatic language understanding.

What carries the argument

The machinery rests on the identity that, for a multivariate normal sample, the scaled squared Mahalanobis distance of a new point follows a Beta distribution: $n/(n-1)^2 \, d^2(\hat\mu,\hat\Sigma) \sim \text{Beta}(d/2, (n-d)/2)$. This converts class membership into a statistically grounded threshold test (Algorithm 1). The companion loss $L_{\mathrm{MAH},\mu}$ uses the covariance of the target embeddings to make the distance scale-invariant, contrasting each target point with the target mean and with sampled negatives.

What would settle it

Compute squared Mahalanobis distances for held-out target and non-target instances using the fine-tuned d-dimensional embeddings and run a multivariate normality test (e.g., Henze-Zirkler) on the target embeddings. If the target distribution is multimodal or heavy-tailed, the Beta critical value will be miscalibrated, and a nonparametric threshold would either change the reported F1/FPR or confirm the assumption.

Watch

Extended reading notes

Core claim

The paper's central discovery is that pragmatic language detection does not need to be framed as ordinary binary classification against a heterogeneous negative class; it can be treated as a geometric distillation of a well-behaved target manifold. ClaD fine-tunes a small pretrained encoder for one epoch with a Mahalanobis contrast loss that makes target embeddings more compact around their mean and pushes negatives away, then decides membership by comparing a test point's normalized squared Mahalanobis distance against a critical value from a Beta distribution. Across sarcasm, metaphor, and sexism benchmarks, this procedure achieves lower false-positive rates than multitask fine-tuned baselines and performance comparable to LLMs up to two orders of magnitude larger, with especially strong results in low-data regimes.

Load-bearing premise

The load-bearing assumption is that the target class's embeddings are approximately multivariate normal in the full fine-tuned space where Mahalanobis distance is computed; the paper's normality evidence comes from 3-D t-SNE projections of pretrained embeddings, not the operational d-dimensional space.

Editorial extensions

If this is right

A single epoch of ClaD training can replace 3-5 epochs of standard fine-tuning on sarcasm, metaphor, and sexism detection without loss of F1 and with lower false-positive rates.
Models with ~110M parameters can match or surpass LLMs two orders of magnitude larger when both see identical task data for one epoch, suggesting scale is not the only route to performance.
In low-resource regimes (5 or 100 training instances), ClaD outperforms all nine tested LLMs on F1 in most tasks, and matches the lowest false-positive rates.
The Beta-decision rule is the primary driver of the gains: replacing it with a small MLP head drops F1 by 46-81%, and the Mahalanobis loss halves FPR relative to a cosine loss.

Reading between the lines

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

Because the Beta threshold assumes a unimodal Gaussian target, efforts to apply ClaD to other 'rare target vs. everything else' tasks (misinformation, offensive language) should first test multivariate normality of fine-tuned target embeddings; multimodal targets would need a mixture model.
The memory cost of a rolling covariance matrix is the likely scaling bottleneck; online low-rank covariance updates are a testable extension that could make the method viable on very large corpora.
The paper's low-FPR results in sexism detection suggest the paradigm could be evaluated for content moderation where avoiding false positives on innocuous comments is a deployment constraint.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Quick take: this is a genuinely new training/inference combination for pragmatic-language minority-class detection, and the paper deserves a serious referee. The central claim is that ClaD—a Mahalanobis contrastive loss plus a Beta-distribution decision rule—lets small encoders match or beat much larger LLMs with one epoch of fine-tuning. That claim might be true, but the evidence as presented is missing the variance information needed to know.

What's new: the integration. Mahalanobis distance, contrastive losses, and Wilks' Beta result are all known, but putting them together as class distillation for tasks like sexism, metaphor, and sarcasm detection is new as far as I can tell. The paper does a good job motivating the problem with the "minority class from heterogeneous background" framing. The ablations are convincing that both components matter: replacing LMAH with a cosine loss or the Beta-decision with an MLP drops F1 sharply. Appendix D's comparison with one-class baselines also helps position the method. The limitations section is candid about the normality assumption and the memory cost of maintaining the covariance matrix.

The soft spots come in two flavors. First, the theoretical premise: the normality of the target class is shown on 3-D t-SNE coordinates, not in the d-dimensional space where Mahalanobis distance is actually computed. The authors say other reductions give similar results, but that's still a proxy. However, this is less damaging than it looks because the Beta threshold v_beta is tuned on development data (footnote 4). That free parameter can absorb some miscalibration, so the algorithm doesn't stand or fall on exact normality.

Second, and more importantly, the empirical headline. The reported numbers appear to come from single runs, no code is released, and the sole decision parameter is dev-tuned with the best test result reported without any uncertainty estimate. OPT is excluded post hoc from Figure 5. That is a recipe for mistaking seed noise for signal. The large ablation margins (e.g., sarcasm F1 0.885 vs 0.620) are probably real, but the smaller gaps against LLMs like Llama3 and Qwen2 could plausibly flip with different seeds. A multi-seed stability analysis with confidence intervals, plus disclosed threshold tuning, would make the central claim convincing.

I'd bring this to a reading group. I'd cite it in work on efficient semantic detection if the authors close the reproducibility gap. It's not circular in the frustrating sense—the loss compacts the target class, which supports the Beta rule, but that's a designed feedback loop, not a procedural error. Yes, send it to peer review, and ask the authors for seeds, code, and a clear account of v_beta calibration.

Referee Report

4 major / 6 minor

Summary. The paper proposes Class Distillation (ClaD), a training and inference paradigm for binary tasks in which the target class is small and the non-target class is highly heterogeneous. During training, a sentence encoder is fine-tuned with a Mahalanobis-contrast loss that pulls target instances toward the target mean and pushes negatives away; at inference, a Mahalanobis Beta-decision rule compares the normalized squared Mahalanobis distance of a test point to a Beta-distribution critical value. The approach is motivated by normality analyses of pretrained BERT and SimCSE embeddings in Section 3.1, and is evaluated on sarcasm detection (Sarcasm Headlines), metaphor detection (TroFi), and sexism detection (CMSB) against encoder baselines, nine LLMs in various low-resource settings, and anomaly-detection baselines. The paper claims that ClaD matches or outperforms competitive baselines and several LLMs with orders of magnitude fewer parameters and only one training epoch.

Significance. If the central empirical claims hold, the paper offers a useful efficiency result: a small encoder trained for one epoch with a geometry-aware loss can be competitive with, or better than, much larger LLMs on three pragmatic language tasks, with particularly low false-positive rates. The paper's strengths include the cross-task evaluation, the explicit ablation separating the loss and decision components, the appendix comparisons with anomaly detectors, and the honest discussion of limitations. However, the empirical claims currently rest on single-run point estimates without uncertainty quantification, and the theoretical justification depends on a normality analysis performed in a projected space rather than in the space where the decision rule is applied. The significance of the result is therefore conditional on additional robustness evidence.

major comments (4)

[Section 3.1 and Section 4.2 (Table 2, Eq. (4))] The normality evidence motivating Eq. (4) is collected on three-dimensional t-SNE coordinates of pretrained BERT and SimCSE embeddings, whereas Algorithm 1 computes Mahalanobis distances in the d-dimensional, fine-tuned representation space. t-SNE is a nonlinear projection that does not preserve Mahalanobis geometry, and the raw HZ and AD statistics in Table 2 are reported without p-values, critical values, or a stated tolerance for what counts as approximately normal. The paper therefore does not establish that the target class is approximately multivariate normal in the space where the Beta decision rule is actually applied. Moreover, no normality check is reported after training with L_MAH,mu, even though the Limitations section says training is expected to nudge embeddings toward a normal manifold; this makes the justification self-reinforcing rather than independently verified. Please either provide evidence of Beta calibration in the actual embedding space or explicitly treat v_beta as an empirically calibrated threshold and remove the distributional claim from the justification.
[Section 5, Table 3, Figures 3-5] The headline comparisons are reported as single-run point estimates. There is no seed information, no error bars, and no confidence intervals for ClaD or for the baselines, and the only bootstrap analysis in the paper (footnote 8) is applied to XLNet rather than to ClaD. Because the central efficiency claim depends on comparative margins, for example the sarcasm F1 gap of 0.885 vs. 0.620 in Table 3 and the gaps over Llama3 and Qwen2 in Figure 5, it is unclear whether the reported advantages exceed seed-to-seed variation. A multi-seed evaluation with means and standard deviations or bootstrap intervals should be reported for the main tables and figures.
[Section 4.2, footnote 4] The critical value v_beta is chosen on development data, so the decision algorithm is not actually setting its threshold from the Beta distribution; the theoretical calibration is replaced by a tuned hyperparameter. This weakens the claim that the decision rule is interpretable and optimized for class separation in a parameter-free sense. Please report the selected v_beta values per task, include a sensitivity analysis around the threshold, and separate development selection from test evaluation.
[Section 5.2, Figure 5, footnote 9] OPT is excluded post hoc from the identical-data comparison with the explanation that it markedly underperformed across all tasks. If the paper claims a complete comparison against the suite of LLMs, either OPT should be included in Figure 5 or an a priori, uniformly applied exclusion rule should be stated. As written, the post hoc exclusion makes the LLM comparison appear selective and should be addressed.

minor comments (6)

[Algorithm 1] The Beta distribution is written as Beta(d/2, n-d/2), which is ambiguous; it should be Beta(d/2, (n-d)/2).
[Appendix D] The sentence 'IS performs better on sarcasm detection' should read 'IF' (Isolation Forest) for consistency with the rest of the appendix.
[References] In the Noumir et al. entry, 'Cédue Richard' should be 'Cédric Richard', and in the Kiska entry, 'Regent Univerity Law Review' should be 'Regent University Law Review'.
[Limitations] The word 'resouce' should be 'resource'.
[Reproducibility] No code or model configuration files are provided; given the number of implementation choices, including the sliding-window covariance updates, LoRA settings, and development-set threshold selection, releasing code or a detailed configuration appendix would materially improve reproducibility.
[Section 3.1, footnote 3] The statement that findings generalize to ALBERT and DistilBERT while 'rare outliers (GPT-2 and Phi)' reflect pretraining misalignment is made without presenting the corresponding results; either add the results or soften the claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the Beta-decision rule rests on an external classical theorem, normality is explicitly flagged as an assumption, and no self-citation or fitted-input-as-prediction chain is present.

full rationale

The paper does not exhibit a circular derivation. The Beta-distribution decision rule (Eq. 4, Algorithm 1) is justified by an external classical result (Wilks 1962; Ververidis and Kotropoulos 2008), not by the authors' own prior work. The normality tests in Section 3.1 are used as motivation for choosing Mahalanobis distance; the decision rule is conditional on the stated normality assumption, and the Limitations section explicitly acknowledges that the assumption may fail ('this assumption may not hold universally'). Setting the threshold v_beta on development data (footnote 4) is standard hyperparameter calibration, not a fitted input that is then reported as a prediction. The loss in Eqs. (1)-(2) and the decision algorithm both use Mahalanobis statistics, but the reported claims are test-set F1/FPR against independently trained baselines and LLMs; the ablations compare component choices on the same data, which is evaluation design rather than circularity. No self-citations appear in the reference list, and no load-bearing premise is imported from the authors' prior work. The central claim is therefore self-contained and falsifiable by external benchmarks.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central method rests on the normality of the target class in the fine-tuned embedding space, a domain assumption tested only indirectly on t-SNE projections. The decision threshold v_beta is a fitted parameter. No new entities are postulated; the method uses existing statistical tools in a new combination.

free parameters (1)

Beta quantile threshold v_beta = not reported (tuned on development data)
Footnote 4 states the critical threshold is 'determined based on development data, ensuring optimal calibration for inference.' This is a fitted decision boundary, not a parameter-free derivation.

assumptions (3)

domain assumption Target class embeddings follow a multivariate normal distribution.
Section 4.2 derives the Beta-distributed squared Mahalanobis distance from this; Table 2 only tests normality on 3-D t-SNE coordinates, not the d-dimensional representation used by the loss and decision rule.
standard math Wilks' result relating normalized squared Mahalanobis distance to a Beta distribution.
Eq. (4) is invoked for Algorithm 1; it is standard for normal samples but requires the normality assumption to transfer to the fine-tuned embedding space.
domain assumption The non-target class is too heterogeneous to be modeled, so the loss uses only target-class statistics.
Design assumption stated in Sections 2 and 3; if non-target structure is informative, ignoring it could hurt performance, but the ablation partially tests this by comparing against cosine loss.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Class Distillation with Mahalanobis Contrast: An Efficient Training Paradigm for Pragmatic Language Understanding Tasks." pith.science (2026). https://pith.science/paper/YCM6XI5S

@misc{pith2026250511829,
  author       = {Pith},
  title        = {Pith review of: Class Distillation with Mahalanobis Contrast: An Efficient Training Paradigm for Pragmatic Language Understanding Tasks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YCM6XI5S}},
  note         = {Machine review of arXiv:2505.11829}
}

read the original abstract

Detecting deviant language such as sexism, or nuanced language such as metaphors or sarcasm, is crucial for enhancing the safety, clarity, and interpretation of online social discourse. While existing classifiers deliver strong results on these tasks, they often come with significant computational cost and high data demands. In this work, we propose \textbf{Cla}ss \textbf{D}istillation (ClaD), a novel training paradigm that targets the core challenge: distilling a small, well-defined target class from a highly diverse and heterogeneous background. ClaD integrates two key innovations: (i) a loss function informed by the structural properties of class distributions, based on Mahalanobis distance, and (ii) an interpretable decision algorithm optimized for class separation. Across three benchmark detection tasks -- sexism, metaphor, and sarcasm -- ClaD outperforms competitive baselines, and even with smaller language models and orders of magnitude fewer parameters, achieves performance comparable to several large language models (LLMs). These results demonstrate ClaD as an efficient tool for pragmatic language understanding tasks that require gleaning a small target class from a larger heterogeneous background.

Figures

Figures reproduced from arXiv: 2505.11829 by the authors.

**Figure 2.** Comparison of ClaD across three detection tasks (from top to bottom) – (a) sarcasm, (b) metaphors, [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Comparison of 5-shot evaluation of a suite of nine large language models (left to right): Llama2, Llama3, [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Comparison of ClaD across the three detection tasks against nine large language models (LLMs) in a [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Comparison of ClaD across the three detection tasks against LLMs, with identical training data: all models [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: Q-Q (quantile-quantile) plots to assess the [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

**Figure 7.** Figure 7: Comparison of zero-shot evaluation of a suite of nine large language models (left to right): Llama2, [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

**Figure 8.** Figure 8: Scatter plots of squared Mahalanobis distance for three test datasets before and after training. Red and [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

66 extracted references · 36 canonical work pages

[1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

goodness of fit

Theodore W. Anderson and Donald Darling. 1952. https://api.semanticscholar.org/CorpusID:120541257 Asymptotic theory of certain "goodness of fit" criteria based on stochastic processes . Annals of Mathematical Statistics, 23:193--212

work page 1952
[4]

Angeliki Athanasiadou. 2024. https://doi.org/10.1016/j.lingua.2023.103655 On the margins of figurative thought and language . Lingua, 299:103655

work page arXiv 2024
[5]

Richard Bellman. 1957. Dynamic Programming . Princeton University Press

work page 1957
[6]

Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. https://doi.org/10.1145/3442188.3445922 On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT '21, page 610–623, New York, NY, USA. Association for Computing...

arXiv 2021
[7]

Julia Birke and Anoop Sarkar. 2006. https://aclanthology.org/E06-1042 A Clustering Approach for Nearly Unsupervised Recognition of Nonliteral Language . In 11th Conference of the E uropean Chapter of the Association for Computational Linguistics , pages 329--336, Trento, Italy. Association for Computational Linguistics

work page 2006
[8]

Saugata Bose and Guoxin Su. 2022. Deep one-class hate speech detection model. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 7040--7048

work page 2022

Show all 66 references

[9]

Daniel Cer, Mona Diab, Eneko Agirre, I \ n igo Lopez-Gazpio, and Lucia Specia. 2017. https://doi.org/10.18653/v1/S17-2001 SemEval-2017 Task 1: Semantic Textual Similarity Multilingual and Crosslingual Focused Evaluation . In Proceedings of the 11th International Workshop on Se...

2017 doi
[10]

Tuhin Chakrabarty, Arkadiy Saakyan, Debanjan Ghosh, and Smaranda Muresan. 2022. FLUTE: Figurative Language Understanding through Textual Explanations . arXiv, 2205.12404v3

2022 arXiv
[11]

Raghavendra Chalapathy and Sanjay Chawla. 2019. https://doi.org/10.48550/arXiv.1901.03407 Deep Learning for Anomaly Detection: A Survey . arXiv, 1901.03407

2019 doi arXiv
[12]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://doi.org/10.18653/v1/N19-1423 BERT : Pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North A merican Chapter of the Associat...

2019 doi
[13]

Kawin Ethayarajh. 2019. https://doi.org/10.18653/v1/D19-1006 How contextual are contextualized word representations? C omparing the geometry of BERT , ELM o, and GPT -2 embeddings . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and t...

2019 doi
[14]

Paula Fortuna and S \'e rgio Nunes. 2018. A Survey on Automatic Detection of Hate Speech in Text . ACM Computing Surveys, 51(4):1--30

2018
[15]

Tianyu Gao, Xingcheng Yao, and Danqi Chen. 2021. https://doi.org/10.18653/V1/2021.EMNLP-MAIN.552 SimCSE: Simple Contrastive Learning of Sentence Embeddings . In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 6894--6910. Associatio...

2021 doi
[16]

Mengshi Ge, Rui Mao, and Erik Cambria. 2023. https://doi.org/10.1007/s10462-023-10564-7 A survey on computational metaphor processing techniques: from identification, interpretation, generation to application . Artif. Intell. Rev., 56(2):1829--1895

2023 doi
[17]

Debanjan Ghosh, Avijit Vajpayee, and Smaranda Muresan. 2020. A Report on the 2020 Sarcasm Detection Shared Task

2020
[18]

Hunter Gregory, Steven Li, Pouya Mohammadi, Natalie Tarn, Rachel Draelos, and Cynthia Rudin. 2020. A transformer approach to contextual sarcasm detection in twitter. In Proceedings of the second workshop on figurative language processing, pages 270--275

2020
[19]

Ella Guest, Bertie Vidgen, Alexandros Mittos, Nishanth Sastry, Gareth Tyson, and Helen Margetts. 2021. https://doi.org/10.18653/v1/2021.eacl-main.114 An Expert Annotated Dataset for the Detection of Online Misogyny . In Proceedings of the 16th Conference of the European Chapte...

2021 doi
[20]

Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, et al. 2025. https://doi.org/10.48550/arXiv.2501.12948 DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning . arXiv, 2501.12948

2025 doi arXiv
[21]

Suchin Gururangan, Swabha Swayamdipta, Omer Levy, Roy Schwartz, Samuel Bowman, and Noah A. Smith. 2018. https://doi.org/10.18653/v1/N18-2017 Annotation artifacts in natural language inference data . In Proceedings of the 2018 Conference of the North A merican Chapter of the As...

2018 doi
[22]

Pengcheng He, Xiaodong Liu, Jianfeng Gao, and Weizhu Chen. 2021. https://openreview.net/forum?id=XPZIaotutsD DeBERTa: Decoding-Enhanced BERT With Disentangled Attention . In International Conference on Learning Representations

2021
[23]

Norbert Henze and Bernd Zirkler. 1990. https://api.semanticscholar.org/CorpusID:120328121 A class of invariant consistent tests for multivariate normality . Communications in Statistics-theory and Methods, 19:3595--3617

1990
[24]

Harold Hotelling. 1936. https://doi.org/10.2307/2333955 Relations between two sets of variates . Biometrika, 28(3/4):321 -- 377

1936 doi
[25]

Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. LoRA: Low-Rank Adaptation of Large Language Models . In International Conference on Learning Representations

2022
[26]

Mojan Javaheripi, S \'e bastien Bubeck, Marah Abdin, Jyoti Aneja, Sebastien Bubeck, Caio C \'e sar Teodoro Mendes, Weizhu Chen, Allie Del Giorno, Ronen Eldan, Sivakanth Gopi, et al. 2023. Phi-2: The surprising power of small language models . Microsoft Research Blog

2023
[27]

Kristina Kasparian. 2013. https://doi.org/10.1016/j.jneuroling.2012.07.001 H emispheric differences in figurative language processing: C ontributions of neuroimaging methods and challenges in reconciling current empirical findings . Journal of Neurolinguistics, 26(1):1--21

2013 doi
[28]

Roger Kiska. 2012. Hate Speech: A Comparison between the European Court of Human Rights and the United States Supreme Court Jurisprudence . Regent Univerity Law Review, 25:107

2012
[29]

Amardeep Kumar and Vivek Anand. 2020. Transformers on sarcasm detection with context. In Proceedings of the second workshop on figurative language processing, pages 88--92

2020
[30]

Huiyuan Lai and Malvina Nissim. 2024. https://doi.org/10.1145/3654795 A Survey on Automatic Generation of Figurative Language: From Rule-based Systems to Large Language Models . ACM Comput. Surv., 56(10)

2024 doi
[31]

Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. 2020. https://iclr.cc/virtual_2020/poster_H1eA7AEtvS.html ALBERT: A Lite BERT for Self-supervised Learning of Language Representations . In International Conference on Learning Repres...

2020
[32]

Jens Lemmens, Ben Burtenshaw, Ehsan Lotfi, Ilia Markov, and Walter Daelemans. 2020. Sarcasm detection using an ensemble approach. In proceedings of the second workshop on figurative language processing, pages 264--269

2020
[33]

Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, et al. 2024. DeepSeek-V3 Technical Report . Technical report, DeepSeek-AI

2024
[34]

Fei Tony Liu, Kai Ming Ting, and Zhi-Hua Zhou. 2008. https://doi.org/10.1109/ICDM.2008.17 Isolation Forest . In 2008 Eighth IEEE International Conference on Data Mining, pages 413--422

2008 doi
[35]

Fei Tony Liu, Kai Ming Ting, and Zhi-Hua Zhou. 2010. https://doi.org/10.1007/978-3-642-15883-4_18 On Detecting Clustered Anomalies Using SCiForest . In Machine Learning and Knowledge Discovery in Databases, pages 274--290, Berlin, Heidelberg. Springer Berlin Heidelberg

2010 doi
[36]

Prasanta Chandra Mahalanobis. 1936. On the Generalized Distance in Statistics . Proceedings of the National Institute of Sciences of India, 2(1):49--55

1936
[37]

Rishabh Misra and Prahal Arora. 2019. https://doi.org/10.48550/arXiv.1908.07414 Sarcasm Detection using Hybrid Neural Network . arXiv, 1908.07414

2019 doi arXiv
[38]

Rishabh Misra and Prahal Arora. 2023. https://doi.org/10.1016/j.aiopen.2023.01.001 Sarcasm detection using news headlines dataset . AI Open, 4:13--18

2023 doi
[39]

Zineb Noumir, Paul Honeine, and C\'edue Richard. 2012. https://doi.org/10.1109/ISIT.2012.6283685 On simple one-class classification methods . In 2012 IEEE International Symposium on Information Theory Proceedings, pages 2022--2026

2012
[40]

Shereen Oraby, Vrindavan Harrison, Lena Reed, Ernesto Hernandez, Ellen Riloff, and Marilyn Walker. 2016. https://doi.org/10.18653/v1/W16-3604 Creating and Characterizing a Diverse Corpus of Sarcasm in Dialogue . In Proceedings of the 17th Annual Meeting of the Special Interest...

2016 doi
[41]

Karl Pearson. 1901. https://doi.org/10.1080/14786440109462720 On Lines and Planes of Closest Fit to Systems of Points in Space . Philosophical Magazine, 2(11):559 -- 572

1901 doi
[42]

Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language Models are Unsupervised Multitask Learners . OpenAI blog, 1(8):9

2019
[43]

Nils Reimers and Iryna Gurevych. 2019. https://doi.org/10.18653/V1/D19-1410 Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on ...

2019 doi
[44]

Ellen Riloff, Ashequl Qadir, Prafulla Surve, Lalindra De Silva, Nathan Gilbert, and Ruihong Huang. 2013. https://aclanthology.org/D13-1066 Sarcasm as Contrast between a Positive Sentiment and Negative Situation . In Proceedings of the 2013 Conference on Empirical Methods in Na...

2013
[45]

Anna Rogers, Olga Kovaleva, and Anna Rumshisky. 2021. https://doi.org/10.1162/tacl_a_00349 A Primer in BERTology: What We Know About How BERT Works . Transactions of the Association for Computational Linguistics, 8:842--866

2021 doi
[46]

Michel Rosenfeld. 2002. Hate Speech in Constitutional Jurisprudence: A Comparative Analysis . Cardozo Law Review, 24:1523

2002
[47]

o rnitz, Lucas Deecke, Shoaib Ahmed Siddiqui, Robert A. Vandermeulen, Alexander Binder, Emmanuel M \

Lukas Ruff, Nico G \" o rnitz, Lucas Deecke, Shoaib Ahmed Siddiqui, Robert A. Vandermeulen, Alexander Binder, Emmanuel M \" u ller, and Marius Kloft. 2018. http://proceedings.mlr.press/v80/ruff18a.html Deep one-class classification . In Proceedings of the 35th International Co...

2018
[48]

Arkadiy Saakyan, Tuhin Chakrabarty, Debanjan Ghosh, and Smaranda Muresan. 2022. https://doi.org/10.18653/v1/2022.flp-1.26 A Report on the FigLang 2022 Shared Task on Understanding Figurative Language . In Proceedings of the 3rd Workshop on Figurative Language Processing (FLP),...

2022 doi
[49]

Mohammad Sabokrou, Mohammad Khalooei, Mahmood Fathy, and Ehsan Adeli. 2018. https://doi.org/10.1109/CVPR.2018.00356 Adversarially learned one-class classifier for novelty detection . In 2018 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2018, Salt Lake City,...

2018
[50]

Niloofar Safi Samghabadi, Parth Patwa, Srinivas PYKL, Prerana Mukherjee, Amitava Das, and Thamar Solorio. 2020. https://aclanthology.org/2020.trac-1.20 Aggression and Misogyny Detection using BERT: A Multi-Task Approach . In Proceedings of the Second Workshop on Trolling, Aggr...

2020
[51]

Call me sexist, but

Mattia Samory, Indira Sen, Julian Kohne, Fabian Flöck, and Claudia Wagner. 2021. https://doi.org/10.1609/icwsm.v15i1.18085 “Call me sexist, but...” : Revisiting Sexism Detection Using Psychological Scales and Adversarial Samples . Proceedings of the International AAAI Conferen...

2021 doi
[52]

Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2019. https://www.emc2-ai.org/assets/docs/neurips-19/emc2-neurips19-paper-33.pdf DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter . In 5th Workshop on Energy Efficient Machine Learning ...

2019
[53]

Platt, John Shawe - Taylor, Alexander J

Bernhard Sch \" o lkopf, John C. Platt, John Shawe - Taylor, Alexander J. Smola, and Robert C. Williamson. 2001. https://doi.org/10.1162/089976601750264965 Estimating the support of a high-dimensional distribution . Neural Comput., 13(7):1443--1471

2001 doi
[54]

Xiaofei Sun, Xiaoya Li, Jiwei Li, Fei Wu, Shangwei Guo, Tianwei Zhang, and Guoyin Wang. 2023. https://doi.org/10.18653/V1/2023.FINDINGS-EMNLP.603 Text classification via large language models . In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore...

2023 doi
[55]

William Timkey and Marten van Schijndel. 2021. https://doi.org/10.18653/v1/2021.emnlp-main.372 All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality . In Proceedings of the 2021 Conference on Empirical Methods in Natural Languag...

2021 doi
[56]

Laurens van der Maaten and Geoffrey Hinton. 2008. Visualizing Data Using t-SNE . Journal of Machine Learning Research, 9:2579 -- 2605

2008
[57]

Dimitrios Ververidis and Constantine Kotropoulos. 2008. Gaussian mixture modeling by exploiting the mahalanobis distance. IEEE transactions on signal processing, 56(7):2797--2811

2008
[58]

Lennart Wachowiak and Dagmar Gromann. 2023. https://doi.org/10.18653/v1/2023.acl-long.58 Does GPT -3 grasp metaphors? identifying metaphor mappings with generative language models . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Vol...

2023 doi
[59]

Hai Wan, Jinxia Lin, Jianfeng Du, Dawei Shen, and Manrong Zhang. 2021. https://doi.org/10.18653/v1/2021.findings-acl.173 Enhancing Metaphor Detection by Gloss-based Interpretations . In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 1971--198...

2021 doi
[60]

Mingyu Wan, Kathleen Ahrens, Emmanuele Chersoni, Menghan Jiang, Qi Su, Rong Xiang, and Chu-Ren Huang. 2020. Using conceptual norms for metaphor detection. In Proceedings of the Second Workshop on Figurative Language Processing, pages 104--109

2020
[61]

Zhiqiang Wang, Yiran Pang, and Yanbin Lin. 2023. https://doi.org/10.48550/ARXIV.2312.01044 Large language models are zero-shot text classifiers . CoRR, abs/2312.01044

2023 doi arXiv
[62]

Samuel S Wilks. 1962. Mathematical statistics. a wiley publication in mathematical statistics john wiley & sons. Inc., New York-London, pages 0173--45805

1962
[63]

Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019. https://proceedings.neurips.cc/paper_files/paper/2019/file/dc6a7e655d7e5840e66733e9ee67cc69-Paper.pdf XLNet: Generalized Autoregressive Pretraining for Language Understanding . In...

2019
[64]

Wenjie Yin and Arkaitz Zubiaga. 2021. Towards generalisable hate speech detection: a review on obstacles and solutions. PeerJ Computer Science, 7:e598

2021
[65]

Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, Todor Mihaylov, Myle Ott, Sam Shleifer, Kurt Shuster, Daniel Simig, Punit Singh Koura, Anjali Sridhar, Tianlu Wang, and Luke Zettlemoyer. 2...

2022 arXiv
[66]

Xianbing Zhou, Yang Yong, Xiaochao Fan, Ge Ren, Yunfeng Song, Yufeng Diao, Liang Yang, and Hongfei Lin. 2021. Hate speech detection based on sentiment knowledge sharing. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Int...

2021

Pith tools

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

[1] [1]

online" 'onlinestring :=

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

goodness of fit

Theodore W. Anderson and Donald Darling. 1952. https://api.semanticscholar.org/CorpusID:120541257 Asymptotic theory of certain "goodness of fit" criteria based on stochastic processes . Annals of Mathematical Statistics, 23:193--212

work page 1952

[4] [4]

Angeliki Athanasiadou. 2024. https://doi.org/10.1016/j.lingua.2023.103655 On the margins of figurative thought and language . Lingua, 299:103655

work page arXiv 2024

[5] [5]

Richard Bellman. 1957. Dynamic Programming . Princeton University Press

work page 1957

[6] [6]

Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. https://doi.org/10.1145/3442188.3445922 On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT '21, page 610–623, New York, NY, USA. Association for Computing...

arXiv 2021

[7] [7]

Julia Birke and Anoop Sarkar. 2006. https://aclanthology.org/E06-1042 A Clustering Approach for Nearly Unsupervised Recognition of Nonliteral Language . In 11th Conference of the E uropean Chapter of the Association for Computational Linguistics , pages 329--336, Trento, Italy. Association for Computational Linguistics

work page 2006

[8] [8]

Saugata Bose and Guoxin Su. 2022. Deep one-class hate speech detection model. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 7040--7048

work page 2022