Pith. sign in

REVIEW 2 major objections 5 minor 37 references

Bias Analysis and Mitigation through Protected Attribute Detection and Regard Classification

T0 review · 2 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A two-step pipeline—protected attribute detection followed by regard classification—can audit and mitigate social bias in Common Crawl, the largest pretraining corpus.

desk verdict A useful corpus-auditing method with a solid SeeGULL check, but the mitigation claim rests on LLM labels that are never validated against humans. read the letter →

arxiv 2504.14212 v3 pith:R3VVBEZO submitted 2025-04-19 cs.CL

Takuma Udagawa , Yang Zhao , Hiroshi Kanayama , Bishwaranjan Bhattacharjee This is my paper · ORCID
classification cs.CL
keywords socialbiasmitigationprotectedattributesregardclassificationwordsensedisambiguationCommonCrawlpretrainingcorpusstereotypeanalysis
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 argues that a scalable two-step annotation pipeline—keyword-plus-gloss word sense disambiguation to detect protected attributes, then regard classification to label the language around each attribute as positive, neutral, or negative—can surface social biases hidden in Common Crawl, the dominant pretraining corpus. Applied to millions of sentences, the pipeline finds systematic regard imbalances, such as "white" appearing with negative polarity far more often than other race/ethnicity attributes. The paper then shows that downsampling negative-regard sentences until no attribute exceeds a 1% negative ratio sharply reduces offensive word associations: the conditional probability of "supremacist" given "white" falls to 19% of its original value and "terrorist" given "arab" to 46%, while neutral or positive associations are largely preserved. If correct, this gives pretraining-data curators a practical way to measure and reduce demographic stereotyping before models are trained, rather than only patching model outputs afterward.

What carries the argument

The load-bearing machinery is the two-stage annotation pipeline. Protected attribute detection first finds mentions of 97 keywords in 10 demographic classes (race/ethnicity, nationality, religion, disability, and others), using Gloss-BERT-style word sense disambiguation: a lightweight RoBERTa classifier decides, for each keyword occurrence, whether the word is used in the sense defined by a hand-crafted gloss (e.g., "blind" as visual impairment, not "blind date"). Regard classification then labels the polarity toward the detected attribute as positive, neutral, or negative, using a RoBERTa classifier distilled from an instruction-tuned Mixtral teacher. The bias score combines both stages by taking, for each word and attribute, the minimum of the word's frequency bias and its regard bias, so that a word only ranks high when it both co-occurs with the attribute and does so with a specific polarity. The mitigation intervention is the distributional consequence of that score: downsample negative-regard sentences until the negative ratio for every attribute is at most 1%.

What would settle it

Have human annotators label a random sample of the same sentences that the pipeline classified as negative toward "white" or "arab" people, and compare the resulting regard distributions; if humans assign substantially fewer negative labels (or a different ordering across demographics), the reported imbalances and the 19%/18%/46% reductions would not survive. Re-running the downsampling with human labels would settle whether the mitigation is an artifact of the teacher's polarity judgments.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that regard-aware co-occurrence statistics, not raw frequency, are the right lens for social-bias auditing of pretraining text. It defines a frequency+regard bias score—the minimum of the frequency bias $p(w|a)/\mathbb{E}_{a\in A}p(w|a)$ and the regard bias $p(r|w,a)/\mathbb{E}_{r\in R}p(r|w,a)$—and shows that sorting words by this score recovers known stereotypes from a geo-cultural stereotype benchmark at higher recall than frequency alone (e.g., negative-stereotype recall at rank 200 rises from 13.97% to 33.09%). It further reports that Common Crawl's regard distributions are imbalanced across demographics, with "white" receiving over 20% negative regard in the race/ethnicity class, and that capping negative regard at 1% by downsampling reduces $p(\text{supremacist}|\text{white})$ to 19%, $p(\text{racists}|\text{white})$ to 18%, and $p(\text{terrorist}|\text{arab})$ to 46% of their original values, while neutral or positive associations stay effectively unchanged.

Load-bearing premise

The entire analysis and mitigation rest on the assumption that the regard labels produced by the teacher model are an accurate proxy for human judgments of positive, neutral, and negative language toward each demographic group; the paper verifies consistency between two teacher runs but does not compare regard labels against human annotation.

Editorial extensions

If this is right

  • Pretraining corpora can be audited at scale for demographic regard imbalance without manual annotation, because the two distilled RoBERTa classifiers run efficiently over hundreds of thousands of sentences.
  • Rebalancing regard distributions is a concrete mitigation: keeping neutral and positive descriptions of a group while thinning negative ones removes the statistical signal behind offensive stereotypes rather than merely suppressing them at generation time.
  • Regard-aware scores identify stereotypes that pure co-occurrence misses, such as "terrorist"–"arab" and "supremacist"–"white", and align better with a stereotype benchmark.
  • The approach can target problematic associations while preserving benign ones—for example, retaining a sentence denying that all white people are racists while downsampling one describing a racist attack.

Reading between the lines

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

  • The paper leaves open whether the same balancing would change what a model actually learns; a direct test would pretrain a small language model on the original and the regard-balanced subsets and compare downstream stereotype measures.
  • Because the regard labels come from a single teacher model, the whole analysis inherits that teacher's polarity judgments; if those judgments are themselves biased, the measured imbalances and reductions describe the teacher's worldview more than the corpus.
  • The same pipeline could be pointed at any large web-crawled corpus or at finer-grained attributes within a demographic class, and the 1% negative-regard cap is an arbitrary threshold—the paper does not explore where the reduction saturates.
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

2 major / 5 minor

Summary. The paper proposes a scalable two-stage annotation pipeline for auditing social bias in large pretraining corpora: (1) protected-attribute detection via keyword matching plus Gloss-BERT-style word sense disambiguation, and (2) regard classification into positive/neutral/negative. Both components are trained on labels distilled from Mixtral models and applied to a subset of Common Crawl. The authors introduce a frequency+regard bias score (Eq. 2) and show it aligns better with SeeGULL stereotypes than frequency-only analysis. They then demonstrate a mitigation intervention that downsamples sentences with negative regard to cap the negative ratio at 1%, reporting large reductions in p('supremacist'|'white'), p('racists'|'white'), and p('terrorist'|'arab'). The paper includes a candid Limitations section.

Significance. If the pipeline is reliable, it addresses a real problem: auditing massive web-crawled corpora for social bias in a scalable, interpretable way. The WSD component with hand-crafted attribute glosses is a sensible adaptation of GlossBERT, and the external validation against SeeGULL in Appendix B is encouraging evidence that the regard-aware score captures known stereotypes better than frequency alone. The paper ships a clear method, honest limitations, and illustrative examples. However, the absence of human validation for the regard labels and the self-referential nature of the mitigation evaluation substantially limit the strength of the central claims. The contribution is promising but currently under-verified.

major comments (2)
  1. [§2.2–§2.3] The regard classifier's reported F1 (0.91 micro, 0.82 macro) is computed against a test set annotated by Mixtral-8x7B and Mixtral-8x22B, not by humans. This measures distillation fidelity, not the validity of the polarity labels. The Limitations correctly state that the teacher 'may be biased in some undesirable ways.' Because the regard labels are used throughout the analysis (Table 2, Figure 2, Tables 7–8) and in the mitigation procedure (§3.2), a systematic polarity bias in Mixtral would propagate to all main results. The authors should add a human-annotated regard test set—even a few hundred sentences per attribute—and report agreement, in the same way they already provide human validation for the WSD stage. Without this, the central analysis claim is not independently verified.
  2. [§3.2] The mitigation evaluation is circular: the same regard classifier that selects negative-regard sentences for downsampling is used to compute the reduction in p(w|a). If 'supremacist' co-occurs with 'white' mostly in sentences that the classifier labels negative, deleting those sentences must reduce p('supremacist'|'white') regardless of whether the label is correct. The reported reductions (19%, 18%, 46%, 26%) are therefore partly a mechanical consequence of the filtering rule. The paper states in the Limitations that no full pretraining ablation was conducted. To make the mitigation claim load-bearing, the authors should evaluate the balanced corpus with an independent measure—for example, a different regard classifier, human judgments, or the SeeGULL-based alignment from §3.1—or explicitly reframe the section as a data-cleaning demonstration rather than a mitigation result.
minor comments (5)
  1. [§3.1, Eq. (2)] The min operation over two ratios with different scales is a heuristic; footnote 6 acknowledges the scale mismatch but the paper does not test sensitivity to alternative combinations (e.g., product or weighted sum). A brief robustness check would increase confidence in the bias score.
  2. [§3.2] The 1% cap on negative regard is arbitrary, and the paper does not report how much data is removed for each attribute or how sensitive the reductions are to the cap value.
  3. [Appendix B, Table 8] The recall@k values (e.g., 2.67, 5.01) are presented as decimals without a stated unit; if they are percentages, they should be marked as such.
  4. [Limitations] There is a typo in the Limitations section: 'regard classier' should be 'regard classifier.'
  5. [General] The paper does not release code, the attribute taxonomy, the trained models, or the annotated corpus, which would make the pipeline easier to reproduce and extend.

Circularity Check

1 steps flagged · score 6.0 of 10

Mitigation result is self-confirming: the reported p(w|a) drops are forced by downsampling sentences the same regard classifier labeled negative.

  1. self definitional [Section 3.2 (Bias Mitigation), using the regard labels from Section 2.2 and the negative-bias definition in eq. (2) of Section 3.1]
    "we ensure the negative regard ratio to be at most 1% for all protected attributes by downsampling negative regard sentences. Then, we compute the relative reduction in the conditional probability p(w|a), where w is a negatively biased word towards a. ... Through this intervention, we confirmed p(w|a) can be dramatically reduced, e.g. down to 19% and 18% for w = “supremacist” and “racists” toward a = “white”, respectively."

    A “negatively biased word” is defined by eq. (2) using the regard labels r=negative produced by the same RoBERTa-based regard classifier (trained on Mixtral-8x7B judgments) that selects which sentences to downsample. Removing exactly those negative-labeled sentences and recomputing p(w|a) on the survivors guarantees that any word whose co-occurrence with a is concentrated in negative-labeled contexts will show a reduced probability; the 19%/18%/46% residuals merely reflect the share of non-negative occurrences that were not removed. The accompanying statement that neutral/positive associations “remain unchanged or slightly increase” is likewise the arithmetic effect of shrinking the denominator after deleting negative sentences.

full rationale

The bias analysis component is not circular: eqs. (1)-(2) are anchored to the external SeeGULL stereotype benchmark (Appendix B, Table 8), and the protected-attribute WSD stage has human-agreement checks (Section 2.3). The circularity is concentrated in the mitigation demonstration (Section 3.2). The downsampling filter removes sentences labeled negative by the regard classifier, and the reported success metric is the drop in p(w|a) for words that eq. (2) identifies as negatively biased using those same labels; such drops are forced by the filtering arithmetic rather than being an independent empirical finding. This is compounded by the absence of human-validated regard labels: the F1=0.91/0.82 in Section 2.3 measures distillation fidelity to Mixtral, not agreement with human judgments, and the Limitations explicitly warn that Mixtral “may be biased in some undesirable ways.” Thus the mitigation loop is closed inside the teacher model’s polarity judgments. Because the analysis portion has external benchmark support and no self-citation chain is load-bearing, the paper is only partially circular rather than wholly reducible to its inputs.

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

The pipeline depends on Mixtral-generated labels for both WSD and regard, and on several hand-chosen thresholds (1% negative cap, 20K vocabulary intersection, sentence length filter). No new physical or mathematical entities are introduced; the bias score is a formula, not an entity.

free parameters (4)
  • negative_regard_cap = 1%
    Used in Section 3.2 to decide how much to downsample negative-regard sentences; the paper does not vary this threshold or report sensitivity analysis.
  • vocabulary_cutoff = 20K words per attribute, intersected
    Appendix B defines the vocabulary V as the intersection of the 20K most frequent words per attribute; this hand-chosen cutoff affects which words appear in the bias tables.
  • sentence_length_filter = 16 to 128 tokens
    Section 3 filters sentences by length using NLTK; the bounds are arbitrary and could affect attribute detection coverage.
  • regard_training_sample_size = 50K sentences per keyword
    Section 2.2 uses 50K Common Crawl sentences per keyword to synthesize regard labels; this is a budget choice, not an optimized value.
assumptions (5)
  • ad hoc to paper Regard distributions should be balanced across protected attributes, with negative regard capped at 1%.
    Stated in Section 3.2 as 'Arguably, a fair pretraining dataset should maintain balanced regard distributions'; the 1% cap is a subjective fairness target, not derived from external constraints.
  • domain assumption Mixtral-8x7B and Mixtral-8x22B provide reliable regard and WSD annotations.
    Training data for both Gloss-RoBERTa and the regard classifier are generated by Mixtral models; the paper relies on these labels throughout and acknowledges in Limitations that the teacher 'may be biased.'
  • domain assumption SeeGULL stereotype annotations are a valid ground truth for positive and negative stereotypes.
    Used in Appendix B to compute recall@k for the bias scores; assumes SeeGULL's offensiveness scores reflect real-world stereotypes.
  • ad hoc to paper The min operation and the unweighted combination in Equation (2) yield a meaningful bias score.
    Equation (2) takes min(frequency bias, regard bias) and ignores scale differences; the paper notes the scale can be adjusted 'if necessary' but gives no justification for the min aggregation.
  • standard math Standard probability calculus.
    Equations (1) and (2) use conditional probability definitions; assumed without proof.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bias Analysis and Mitigation through Protected Attribute Detection and Regard Classification." pith.science (2026). https://pith.science/paper/R3VVBEZO

@misc{pith2026250414212,
  author       = {Pith},
  title        = {Pith review of: Bias Analysis and Mitigation through Protected Attribute Detection and Regard Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R3VVBEZO}},
  note         = {Machine review of arXiv:2504.14212}
}
read the original abstract

Large language models (LLMs) acquire general linguistic knowledge from massive-scale pretraining. However, pretraining data mainly comprised of web-crawled texts contain undesirable social biases which can be perpetuated or even amplified by LLMs. In this study, we propose an efficient yet effective annotation pipeline to investigate social biases in the pretraining corpora. Our pipeline consists of protected attribute detection to identify diverse demographics, followed by regard classification to analyze the language polarity towards each attribute. Through our experiments, we demonstrate the effect of our bias analysis and mitigation measures, focusing on Common Crawl as the most representative pretraining corpus.

Figures

Figures reproduced from arXiv: 2504.14212 by the authors.

Figure 1
Figure 1. An overview of our annotation pipeline consisting of (1) [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Regard distributions for A = race/ethnicity. score: min p(w|a) J p(w|a)Ka∈A , p(r|w, a) J p(r|w, a)Kr∈R  (2) Here, the left term represents frequency bias (eq. (1)) and the right term regard bias of when a word w and attribute a co-occur.6 For instance, if w = supremacist and a = white tend to co-occur with negative regard, the score increases for r = negative. This way, we can analyze not only whether w and a co-o… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 5 canonical work pages

  1. [1]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. 2023. GPT -4 technical report. arXiv preprint arXiv:2303.08774

    arXiv 2023

  2. [2]

    Stefan Baack. 2024. A critical analysis of the largest source for generative ai training data: Common crawl. In The 2024 ACM Conference on Fairness, Accountability, and Transparency, pages 2199--2208

    work page 2024

  3. [3]

    Steven Bird. 2006. NLTK : the natural language toolkit. In Proceedings of the COLING/ACL 2006 Interactive Presentation Sessions, pages 69--72

    work page 2006

  4. [4]

    Su Lin Blodgett, Gilsinia Lopez, Alexandra Olteanu, Robert Sim, and Hanna Wallach. 2021. https://doi.org/10.18653/v1/2021.acl-long.81 Stereotyping N orwegian salmon: An inventory of pitfalls in fairness benchmark datasets . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conferenc...

    doi:10.18653/v1/2021.acl-long.81 2021

  5. [5]

    Shikha Bordia and Samuel R. Bowman. 2019. https://doi.org/10.18653/v1/N19-3002 Identifying and reducing gender bias in word-level language models . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Student Research Workshop , pages 7--15, Minneapolis, Minnesota. Association for Computati...

    doi:10.18653/v1/n19-3002 2019

  6. [6]

    Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, et al. 2023. PaLM : Scaling language modeling with pathways. Journal of Machine Learning Research, 24(240):1--113

    work page 2023

  7. [7]

    Jwala Dhamala, Tony Sun, Varun Kumar, Satyapriya Krishna, Yada Pruksachatkun, Kai-Wei Chang, and Rahul Gupta. 2021. https://doi.org/10.1145/3442188.3445924 BOLD : Dataset and metrics for measuring biases in open-ended language generation . In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, FAccT '21, page 862–872, New...

    arXiv 2021

  8. [8]

    Jesse Dodge, Maarten Sap, Ana Marasovi \'c , William Agnew, Gabriel Ilharco, Dirk Groeneveld, Margaret Mitchell, and Matt Gardner. 2021. Documenting large webtext corpora: A case study on the colossal clean crawled corpus. arXiv preprint arXiv:2104.08758

    arXiv 2021

Show all 37 references
  1. [9]

    Yi Dong, Ronghui Mu, Yanghao Zhang, Siqi Sun, Tianle Zhang, Changshun Wu, Gaojie Jin, Yi Qi, Jinwei Hu, Jie Meng, et al. 2024. Safeguarding large language models: A survey. arXiv preprint arXiv:2406.02622

    2024 arXiv

  2. [10]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The Llama 3 herd of models. arXiv preprint arXiv:2407.21783

    2024 arXiv

  3. [11]

    David Esiobu, Xiaoqing Tan, Saghar Hosseini, Megan Ung, Yuchen Zhang, Jude Fernandes, Jane Dwivedi-Yu, Eleonora Presani, Adina Williams, and Eric Smith. 2023. https://doi.org/10.18653/v1/2023.emnlp-main.230 ROBBIE : Robust bias evaluation of large generative language models . ...

    2023 doi

  4. [12]

    Shangbin Feng, Chan Young Park, Yuhan Liu, and Yulia Tsvetkov. 2023. https://doi.org/10.18653/v1/2023.acl-long.656 From pretraining data to language models to downstream tasks: Tracking the trails of political biases leading to unfair NLP models . In Proceedings of the 61st An...

    2023 doi

  5. [13]

    Deep Ganguli, Amanda Askell, Nicholas Schiefer, Thomas I Liao, Kamil \.e Luko s i \=u t \.e , Anna Chen, Anna Goldie, Azalia Mirhoseini, Catherine Olsson, Danny Hernandez, et al. 2023. The capacity for moral self-correction in large language models. arXiv preprint arXiv:2302.07459

    2023 arXiv

  6. [14]

    Samuel Gehman, Suchin Gururangan, Maarten Sap, Yejin Choi, and Noah A. Smith. 2020. https://doi.org/10.18653/v1/2020.findings-emnlp.301 R eal T oxicity P rompts: Evaluating neural toxic degeneration in language models . In Findings of the Association for Computational Linguist...

    2020 doi

  7. [15]

    Hila Gonen and Yoav Goldberg. 2019. https://doi.org/10.18653/v1/N19-1061 Lipstick on a pig: D ebiasing methods cover up systematic gender biases in word embeddings but do not remove them . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association ...

    2019 doi

  8. [16]

    Luyao Huang, Chi Sun, Xipeng Qiu, and Xuanjing Huang. 2019. https://doi.org/10.18653/v1/D19-1355 G loss BERT : BERT for word sense disambiguation with gloss knowledge . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th Intern...

    2019 doi

  9. [17]

    Hakan Inan, Kartikeya Upasani, Jianfeng Chi, Rashi Rungta, Krithika Iyer, Yuning Mao, Michael Tontchev, Qing Hu, Brian Fuller, Davide Testuggine, et al. 2023. Llama guard: Llm-based input-output safeguard for human-ai conversations. arXiv preprint arXiv:2312.06674

    2023 arXiv

  10. [18]

    Akshita Jha, Aida Mostafazadeh Davani, Chandan K Reddy, Shachi Dave, Vinodkumar Prabhakaran, and Sunipa Dev. 2023. https://doi.org/10.18653/v1/2023.acl-long.548 S ee GULL : A stereotype benchmark with broad geo-cultural coverage leveraging generative models . In Proceedings of...

    2023 doi

  11. [19]

    Albert Q Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Emma Bou Hanna, Florian Bressand, et al. 2024. Mixtral of experts. arXiv preprint arXiv:2401.04088

    2024 arXiv

  12. [20]

    Bill Yuchen Lin, Abhilasha Ravichander, Ximing Lu, Nouha Dziri, Melanie Sclar, Khyathi Chandu, Chandra Bhagavatula, and Yejin Choi. 2024. https://openreview.net/forum?id=wxJ0eXwwda The unlocking spell on base LLM s: Rethinking alignment via in-context learning . In The Twelfth...

    2024

  13. [21]

    Yinhan Liu. 2019. RoBERTa : A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692, 364

    2019 arXiv

  14. [22]

    Shayne Longpre, Gregory Yauney, Emily Reif, Katherine Lee, Adam Roberts, Barret Zoph, Denny Zhou, Jason Wei, Kevin Robinson, David Mimno, and Daphne Ippolito. 2024. https://doi.org/10.18653/v1/2024.naacl-long.179 A pretrainer ' s guide to training data: Measuring the effects o...

    2024 doi

  15. [23]

    Kaiji Lu, Piotr Mardziel, Fangjing Wu, Preetam Amancharla, and Anupam Datta. 2020. Gender bias in neural natural language processing. Logic, language, and security: essays dedicated to Andre Scedrov on the occasion of his 65th birthday, pages 189--202

    2020

  16. [24]

    Alexandra Luccioni and Joseph Viviano. 2021. https://doi.org/10.18653/v1/2021.acl-short.24 What's in the B ox? A n analysis of undesirable content in the C ommon C rawl corpus . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the ...

    2021 doi

  17. [25]

    George A Miller. 1995. WordNet : A lexical database for E nglish. Communications of the ACM, 38(11):39--41

    1995

  18. [26]

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 3...

    2022

  19. [27]

    Guilherme Penedo, Hynek Kydl \' c ek, Anton Lozhkov, Margaret Mitchell, Colin Raffel, Leandro Von Werra, Thomas Wolf, et al. 2024. The FineWeb datasets: Decanting the web for the finest text data at scale. arXiv preprint arXiv:2406.17557

    2024 arXiv

  20. [28]

    Rebecca Qian, Candace Ross, Jude Fernandes, Eric Michael Smith, Douwe Kiela, and Adina Williams. 2022. https://doi.org/10.18653/v1/2022.emnlp-main.646 Perturbation augmentation for fairer NLP . In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Proc...

    2022 doi

  21. [29]

    Timo Schick, Sahana Udupa, and Hinrich Sch \"u tze. 2021. Self-diagnosis and self-debiasing: A proposal for reducing corpus-based bias in nlp. Transactions of the Association for Computational Linguistics, 9:1408--1424

    2021

  22. [30]

    Patrick Schramowski, Cigdem Turan, Nico Andersen, Constantin A Rothkopf, and Kristian Kersting. 2022. Large pre-trained language models contain human-like biases of what is right and wrong to do. Nature Machine Intelligence, 4(3):258--268

    2022

  23. [31]

    Emily Sheng, Kai-Wei Chang, Prem Natarajan, and Nanyun Peng. 2021. https://doi.org/10.18653/v1/2021.acl-long.330 Societal biases in language generation: Progress and challenges . In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the...

    2021 doi

  24. [32]

    Emily Sheng, Kai-Wei Chang, Premkumar Natarajan, and Nanyun Peng. 2019. https://doi.org/10.18653/v1/D19-1339 The woman worked as a babysitter: On biases in language generation . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9...

    2019 doi

  25. [33]

    I 'm sorry to hear that

    Eric Michael Smith, Melissa Hall, Melanie Kambadur, Eleonora Presani, and Adina Williams. 2022. " I 'm sorry to hear that": Finding new biases in language models with a holistic descriptor dataset. arXiv preprint arXiv:2205.09209

    2022 arXiv

  26. [34]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824--24837

    2022

  27. [35]

    Chunting Zhou, Pengfei Liu, Puxin Xu, Srinivasan Iyer, Jiao Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, Lili Yu, et al. 2024. LIMA : Less is more for alignment. Advances in Neural Information Processing Systems, 36

    2024

  28. [36]

    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...

  29. [37]

    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 gl...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.