Pith. sign in

REVIEW 3 major objections 5 minor 43 references

Profiling Bias in LLMs: Stereotype Dimensions in Contextual Word Embeddings

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that projecting contextual embeddings onto stereotype dimensions from social psychology reveals significant gender bias in all twelve studied LLMs, with most models placing female names at warmth and male names at…

desk verdict Useful empirical extension of stereotype-content profiling to twelve LLMs; the gender asymmetry is robust across models, but the dimension validation is partly circular and the cutoff handling is post hoc. read the letter →

arxiv 2411.16527 v2 pith:HJNFSAAL submitted 2024-11-25 cs.CL

Carolin M. Schuster , Maria-Alexandra Dinisor , Shashwat Ghatiwala , Georg Groh This is my paper · ORCID
classification cs.CL
keywords biasinLLMsstereotypecontentmodelgendercontextualwordembeddingsprofilespolarprojectionwarmthandcompetenceembeddinginterpretability
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 the hidden embedding spaces of large language models can be made legible by projecting them onto stereotype dimensions from social psychology, and that doing so reveals a consistent gendered pattern. Across twelve open LLMs, historically female-associated names project toward the warmth pole and male-associated names toward the competence pole; under gender-neutral generated contexts, every model shows statistically significant differences for names, and most show the same female-warmth/male-competence alignment. The authors present these projections as bias profiles: visual, intuitive descriptions of where social groups sit on dimensions like sociability, morality, ability, and agency, intended for communicating risk to non-specialists and for guiding debiasing. The claim matters because it connects the opaque internal states of LLMs to a well-established human psychology framework, giving bias measurement a theoretical grounding rather than a purely correlational one.

What carries the argument

The carrying mechanism is a polar projection onto stereotype axes, following the SensePolar approach. For each dimension, the averaged contextual embeddings of the seed dictionary's high-pole words ($p_{\mathrm{high}}$) and low-pole words ($p_{\mathrm{low}}$) define a direction vector $a_h = p_{\mathrm{high}} - p_{\mathrm{low}}$; stacked, these form a change-of-basis matrix $A$, and a target term's averaged embedding $x$ is projected as $d = (A^T)^{-1}x$, so the sign of $d$ indicates association with the high or low pole. The warmth and competence axes aggregate the finer dimensions (sociability plus morality, and ability plus agency), while the 7D profile keeps these subdimensions separate. This transformation is what turns opaque high-dimensional embeddings into the two- or seven-number profiles the paper visualizes.

What would settle it

Rebuild the warmth and competence pole vectors from an independently sourced, human-rated antonym list, re-project the same 100 male and female names, and check whether the female-warmth/male-competence separation persists; if it disappears or flips, the bias profile is an artifact of the seed-word choice.

Watch

Extended reading notes

Core claim

The core claim is that the stereotype content model's two organizing dimensions, warmth and competence, are present in the contextual embeddings of current LLMs and that gender-associated names are positioned along them in a way that mirrors human stereotype research. Using gender-non-specific generated contexts, the 100 historically female and male names separate significantly on these axes in all twelve models, with female names associated with warmth and male names with competence in nearly all cases; the two exceptions are one model where both dimensions favor female names and one where the competence difference is not significant. The seven-dimensional profiles add detail: sociability and morality drive the female warmth association in ten or eleven models, while ability, agency, and sometimes status drive the male competence association. The authors further report that these associations are distributed across network layers rather than concentrated in any single layer, and that a smaller set of nine gendered terms shows a reliable warmth bias for the newest Llama models, with the direction of the term differences consistent with the name profiles.

Load-bearing premise

The stereotype axes are validated with test words drawn from the same dictionary that defined the poles, so the measured gender asymmetry could reflect the dictionary's internal word geometry rather than stereotype content that holds independently.

Editorial extensions

If this is right

  • Bias in these models is not only an output-level artifact: it is encoded in the geometry of hidden representations and visible through a simple linear projection.
  • The same stereotype axes can serve as a group-agnostic measurement device across model families, sizes, and generations, without needing examples of the discriminated group in the test data.
  • Choice of context changes the measured profile, so gender-neutral generated contexts provide a baseline of pre-existing bias while domain-specific contexts can probe task-relevant bias.
  • Because stereotype associations persist across layers, mitigation that targets only the final layer would leave most of the representational bias in place.
  • Bias profiles give a non-expert-accessible visualization of representational harm, suitable for model cards and public accountability.

Reading between the lines

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

  • The dimensions' validation is dictionary-internal; a natural extension is to rebuild the axes from an independent, human-rated antonym list and test whether the female-warmth/male-competence asymmetry survives.
  • The method's statistical power comes from large name sets, so profiling individual terms or small gender groups will require many more context samples or a Bayesian treatment to be trustworthy.
  • The lower warmth of nonbinary and transgender terms in newer models is suggestive; with more contexts it could be developed into a quantitative diagnostic for harms beyond binary gender.
  • Because the dimensions come from social psychology rather than from the target group, the same projection could be turned toward ethnicity, age, or other stereotype dimensions without changing the measurement pipeline.
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 / 5 minor

Summary. The paper proposes a method for profiling gender bias in large language models by projecting contextual word embeddings onto stereotype dimensions from the stereotype content model (warmth, competence, and seven granular facets). The authors construct polar axes from seed terms in the Nicolas et al. (2021) dictionary, apply a linear projection to embeddings of gender-associated names and gendered terms, and generate 2D and 7D stereotype profiles for twelve open-source LLMs. They report statistically significant gender differences in the projected warmth and competence values, largely consistent with human stereotype content studies, and analyze how the results vary across context types, model layers, and model families. The paper also makes available code for the method.

Significance. If the stereotype dimensions are accepted as valid, the paper offers a practical, interpretable tool for exposing and communicating representational bias in LLMs, which is valuable for both AI audit and public communication. The authors provide a broad empirical sweep across twelve models, use theory-grounded dictionaries, make their code available, and back the headline gender-bias claim with t-tests on a set of 100 names per gender. The layer-wise and context-type analyses add useful nuance. The main caveat is that the validation of the stereotype axes relies on the same dictionary family that defines the axes, so the construct validity of the dimensions is the least secure part of the argument.

major comments (3)
  1. [Sec. 3.3 and Sec. 4.1] The validation of the stereotype dimensions is partly circular. The polar axes are defined by seed terms from the Nicolas et al. (2021) dictionary, and the direction-prediction task in Sec. 4.1 evaluates projection accuracy on "additional terms from the extended stereotype dictionary," which is the same dictionary, built semi-automatically as synonyms of the seed terms via WordNet. High accuracy on this internal task mostly shows that the projection separates seed words from their near-synonyms; it does not independently establish that the axes correspond to the psychological warmth and competence constructs. This matters because the paper interprets the gender profiles as evidence about SCM-aligned bias. I recommend an external validation, for example predicting human-rated stereotype content for social groups (as in Fraser et al. 2021) or using a dictionary from a different source, or at least adding an explicit statement that the validation is internal consistency rather than independent construct validation.
  2. [Sec. 3.3, Eq. (5)] Equation (5) writes d = (a^T)^{-1} x, but a is an h×d matrix with h=2 or 7 and d equal to the embedding dimension (e.g., 768), so a^T is not square and has no inverse. The authors must state whether they use the Moore-Penrose pseudo-inverse or a least-squares solution, and define the projection accordingly. As written, the formula is mathematically undefined and the projected values are not reproducible without the code's specific numerical routine.
  3. [Sec. 4.1] The mean-centered cutoff for GPT-Neo, GPT2, and the Gemma models is introduced after observing poor accuracy with the zero cutoff. This is a post hoc adjustment that affects the accuracy numbers in Table 2 and the claim that "all models can reasonably discriminate between low and high-labeled terms." The paper should either motivate the cutoff a priori (e.g., because the projection has no built-in centering) or report both zero-cutoff and mean-centered results without presenting the latter as the default. This does not necessarily invalidate the gender-bias t-tests in Sec. 4.2, since a constant shift would not change the between-group difference, but the validation evidence for those models is weakened.
minor comments (5)
  1. [Sec. 3.3] The construction of the warmth and competence directions from subordinate dimensions should be stated explicitly: are the word lists simply concatenated before averaging, or are the subordinate direction vectors averaged? This affects the exact definition of the axes.
  2. [Sec. 4.2] The sentence "For all twelve studied LLMs, we find statistically significant bias for gender-associated names" is followed by the caveat that GPT-Neo shows associations of both dimensions with female names and OLMo has an insignificant competence difference. Please clarify that the claim refers to at least one dimension being significantly different, and quantify the p-value threshold and correction method used for the t-tests.
  3. [Fig. 1 and Fig. 5] There is a typo "Lama-3-8B" in Figure 1 and "T erms" in Figure 3; please fix these spelling errors.
  4. [Sec. 3.2] The number of context examples is set to five, but the paper does not report how sensitive the results are to this choice. A brief analysis or discussion would help assess robustness.
  5. [Sec. 4.1] For the religion dimension, the imbalance between high-labeled (142) and low-labeled (6) additional terms is noted; this makes the reported accuracy less informative. Please also report per-class or balanced accuracy for that dimension.

Circularity Check

1 steps flagged · score 4.0 of 10

Direction-prediction validation is internal to the Nicolas et al. dictionary; gender projections are held out, so circularity is partial.

  1. self definitional [Sec. 3.3 'Polar Projection' (Projection of Additional Terms); validated in Sec. 4.1; dictionary source in Sec. 3.1]
    "To evaluate the consistency of the stereotype dimensions we follow the approach by Fraser et al. (2021) and project additional terms from the extended dictionary by Nicolas et al. (2021) to the stereotype space... For each term, we predict its direction on its assigned dimension by the sign of its polar value... we compare these predictions against the labels in the dictionary. ..."

    Equation 3 defines each stereotype axis a_h as p_high - p_low, where p is the average embedding of the seed words labeled high/low in the Nicolas et al. (2021) dictionary. Section 4.1 then 'predicts' directions for additional terms drawn from the same extended dictionary and scores them against that dictionary's labels. Since the extended terms were collected by semi-automated WordNet synonym expansion around the same seed lists, this task checks internal consistency of one dictionary: it shows that near-synonyms of the seed words fall on the same side of the seed-derived hyperplane. It does not independently establish that the axis captures warmth/competence rather than a shared valence factor, so the Sec.

full rationale

The central empirical claim—consistent warmth/female and competence/male associations for 100 held-out names across twelve LLMs—is not derived from a fitted parameter or from the paper's own prior work. The axes are computed from theory-driven seed terms (Eq. 3), and the gender projections use externally sourced names and neutral templates with t-tests, so the gender differences are genuine measurements given the axes. No load-bearing self-citation occurs: SensePolar, POLAR, and the Nicolas et al. dictionary are external prior work, and the single author self-citation (Schuster and Hegelich 2022) is a related-work pointer. However, the paper's only validation that the projected dimensions are stereotype dimensions, not dictionary-internal geometry, is the direction-prediction task of Sec. 4.1, whose test terms and labels come from the same Nicolas et al. dictionary used to build the poles. That makes the validation self-referential in a mild but real way. The paper's own finding that some models need mean-centered cut-offs to exceed chance further shows the zero-sign threshold is not a stable, theory-independent scale. Because the dictionary itself was human-validated in the source work, this is a partial construct-validity gap rather than a full 'prediction equals input' reduction, leading to a score of 4.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claims rest on borrowed dictionaries and a borrowed projection method. The only newly chosen parameters are design choices (context count, cutoff), and the main assumption with the largest risk is that the generated contexts are neutral and that the dictionary-based projection captures true stereotype content.

free parameters (2)
  • Context example count = 5
    Set to five per term to limit computational time; affects robustness of sense and pole embeddings but is not fitted to the bias claim.
  • Mean-centered direction cutoff = model-dependent mean, not specified numerically
    Used post hoc in Sec 4.1 to improve direction prediction accuracy for GPT-Neo, GPT2, and Gemma; this hand-chosen threshold inflates apparent accuracy and is a free choice.
assumptions (4)
  • domain assumption The stereotype content model and the Nicolas et al. (2021) dictionaries validly operationalize stereotype dimensions.
    The paper relies on these dictionaries as ground truth for constructing the polar space and for evaluating direction prediction (Sec 3.1, 4.1).
  • domain assumption Averaging contextual embeddings across examples and layers and applying the SensePolar projection yields meaningful stereotype dimensions.
    The entire measurement pipeline (Eqs. 1-5) assumes this without independent validation in the paper.
  • domain assumption Gender-non-specific contexts generated by Llama-3-8B-instruct do not introduce systematic gender bias.
    The main profiles use these generated contexts; if the generator itself is biased, the measured gender differences could be partly an artifact of the contexts (Sec 3.2).
  • standard math Standard linear algebra and t-test assumptions.
    Projection via the inverse change-of-basis matrix and significance testing rely on standard statistical assumptions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Profiling Bias in LLMs: Stereotype Dimensions in Contextual Word Embeddings." pith.science (2026). https://pith.science/paper/HJNFSAAL

@misc{pith2026241116527,
  author       = {Pith},
  title        = {Pith review of: Profiling Bias in LLMs: Stereotype Dimensions in Contextual Word Embeddings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HJNFSAAL}},
  note         = {Machine review of arXiv:2411.16527}
}
read the original abstract

Large language models (LLMs) are the foundation of the current successes of artificial intelligence (AI), however, they are unavoidably biased. To effectively communicate the risks and encourage mitigation efforts these models need adequate and intuitive descriptions of their discriminatory properties, appropriate for all audiences of AI. We suggest bias profiles with respect to stereotype dimensions based on dictionaries from social psychology research. Along these dimensions we investigate gender bias in contextual embeddings, across contexts and layers, and generate stereotype profiles for twelve different LLMs, demonstrating their intuition and use case for exposing and visualizing bias.

Figures

Figures reproduced from arXiv: 2411.16527 by the authors.

Figure 1
Figure 1. 7D stereotype profile for Llama-3-8B, revealing differences in embeddings of 100 female and 100 male-associated names. *Statistically significant differences (p<0.05). (Mathew et al., 2020; Kwak et al., 2021; ¸Senel et al., 2022; Engler et al., 2022), enabling new ways to study concepts. Similar to semantic dif￾ferentials (Osgood et al., 1957), this methodology relies on antonyms (e.g. fast vs. slow) or opposing con… view at source ↗
Figure 2
Figure 2. Properties of context examples: Histograms of example counts, numbers of words and positions of dictionary terms within the examples. 3.2 Context Examples As we are working with contextual embeddings the context of the terms becomes a crucial design choice for the study of stereotype dimensions and bias (see also Engler et al. (2022)). Generated Examples: For our main exper￾iments, we generate gender non-specific co… view at source ↗
Figure 3
Figure 3. 2D stereotype profiles for 100 female/male-associated names (left) and 9 female/male gendered terms (right). LW/HW = Low/High Warmth. LC/HC = Low/High Competence. *Dimensions with statistically significant differences (p<0.05). [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: 2D Stereotype profile for Llama-3-8B (see [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: 7D stereotype profiles for 100 female/male-associated names (left) and 9 female/male [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Layerwise visualization of prediction accuracy and gender bias for selected models. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 22 canonical work pages

  1. [1]

    URL: " 'urlintro :=

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

  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]

    Abele, Nicole Hauke, Kim Peters, Eva Louvet, Aleksandra Szymkow, and Yanping Duan

    Andrea E. Abele, Nicole Hauke, Kim Peters, Eva Louvet, Aleksandra Szymkow, and Yanping Duan. 2016. Facets of the fundamental content dimensions: Agency with competence and assertiveness— Communion with warmth and morality. Frontiers in psychology, 7:1810

    work page 2016

  4. [4]

    AI@Meta. 2024 a . Llama 3 model card. URL: https://github.com/meta-llama/llama3 /blob/main/MODEL\_CARD.md

    work page 2024

  5. [5]

    AI@Meta. 2024 b . Llama 3.2 model card. URL: https://github.com/meta-llama/llama-models /blob/main/models/llama3\_2/MODEL\_CARD.md

    work page 2024

  6. [6]

    Jisun An, Haewoon Kwak, and Yong-Yeol Ahn. 2018. https://doi.org/10.18653/v1/P18-1228 S em A xis: A lightweight framework to characterize domain-specific word semantics beyond sentiment . In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2450--2461, Melbourne, Australia. Association f...

    doi:10.18653/v1/p18-1228 2018

  7. [7]

    Sid Black, Leo Gao, Phil Wang, Connor Leahy, and Stella Biderman. 2022. https://doi.org/10.5281/zenodo.5297715 Gpt-neo: Large scale autoregressive language modeling with mesh-tensorflow, 2021 . URL: https://doi. org/10.5281/zenodo, 5297715

    doi:10.5281/zenodo.5297715 2022

  8. [8]

    Su Lin Blodgett, Solon Barocas, Hal Daum \'e III, and Hanna Wallach. 2020. https://doi.org/10.18653/v1/2020.acl-main.485 Language (technology) is power: A critical survey of `` bias '' in NLP . In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5454--5476, Online. Association for Computational Linguistics

    doi:10.18653/v1/2020.acl-main.485 2020

Show all 43 references
  1. [9]

    Bryson, and Arvind Narayanan

    Aylin Caliskan, Joanna J. Bryson, and Arvind Narayanan. 2017. https://doi.org/10.1126/science.aal4230 Semantics derived automatically from language corpora contain human-like biases . Science, 356(6334):183--186

    2017 doi

  2. [10]

    Boxi Cao, Hongyu Lin, Xianpei Han, and Le Sun. 2024. The life cycle of knowledge in big language models: A survey. Machine Intelligence Research, 21(2):217--238

    2024

  3. [11]

    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

  4. [12]

    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

  5. [13]

    Naomi Ellemers. 2017. Morality and the regulation of social behavior: Groups as moral anchors . Psychology Press

    2017

  6. [14]

    Jan Engler, Sandipan Sikdar, Marlene Lutz, and Markus Strohmaier. 2022. https://doi.org/10.18653/v1/2022.findings-emnlp.338 S ense POLAR : Word sense aware interpretability for pre-trained contextual word embeddings . In Findings of the Association for Computational Linguistic...

    2022 doi

  7. [15]

    Susan T Fiske, Amy JC Cuddy, Peter Glick, and Jun Xu. 2002. A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82(6):878--902

    2002

  8. [16]

    Fraser, Isar Nejadgholi, and Svetlana Kiritchenko

    Kathleen C. Fraser, Isar Nejadgholi, and Svetlana Kiritchenko. 2021. https://doi.org/10.18653/v1/2021.acl-long.50 Understanding and countering stereotypes: A computational approach to the stereotype content model . In Proceedings of the 59th Annual Meeting of the Association f...

    2021 doi

  9. [17]

    Isabel O Gallegos, Ryan A Rossi, Joe Barrow, Md Mehrab Tanjim, Sungchul Kim, Franck Dernoncourt, Tong Yu, Ruiyi Zhang, and Nesreen K Ahmed. 2024. Bias and fairness in large language models: A survey. Computational Linguistics, pages 1--79

    2024

  10. [18]

    Geoffrey P. Goodwin. 2015. https://doi.org/10.1177/0963721414550709 Moral Character in Person Perception . Current Directions in Psychological Science, 24(1):38--44

    2015 doi

  11. [19]

    Greenwald, Debbie E

    Anthony G. Greenwald, Debbie E. McGhee, and Jordan LK Schwartz. 1998. https://psycnet.apa.org/record/1998-02892-004 Measuring individual differences in implicit cognition: the implicit association test. Journal of personality and social psychology, 74(6):1464

    1998

  12. [20]

    Dirk Groeneveld, Iz Beltagy, Pete Walsh, Akshita Bhagia, Rodney Kinney, Oyvind Tafjord, Ananya Harsh Jha, Hamish Ivison, Ian Magnusson, Yizhong Wang, Shane Arora, David Atkinson, Russell Authur, Khyathi Chandu, Arman Cohan, Jennifer Dumas, Yanai Elazar, Yuling Gu, Jack Hessel,...

    2024

  13. [21]

    Wei Guo and Aylin Caliskan. 2021. Detecting emergent intersectional biases: Contextualized word embeddings contain a distribution of human-like biases. In Proceedings of the 2021 AAAI / ACM Conference on AI , Ethics , and Society , pages 122--133

    2021

  14. [22]

    Heilman, Aaron S

    Madeline E. Heilman, Aaron S. Wallen, Daniella Fuchs, and Melinda M. Tamkins. 2004. https://psycnet.apa.org/journals/apl/89/3/416/ Penalties for success: reactions to women who succeed at male gender-typed tasks. Journal of applied psychology, 89(3):416

    2004

  15. [23]

    Sullam Jeoung, Yubin Ge, and Jana Diesner. 2023. https://doi.org/10.18653/v1/2023.emnlp-main.752 S tereo M ap: Quantifying the awareness of human-like stereotypes in large language models . In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processi...

    2023 doi

  16. [24]

    Alex Koch, Roland Imhoff, Ron Dotsch, Christian Unkelbach, and Hans Alves. 2016. https://psycnet.apa.org/record/2016-22171-004 The ABC of stereotypes about groups: Agency /socioeconomic success, conservative–progressive beliefs, and communion. Journal of personality and social...

    2016

  17. [25]

    Haewoon Kwak, Jisun An, Elise Jing, and Yong-Yeol Ahn. 2021. FrameAxis : characterizing microframe bias and intensity with word embedding. PeerJ Computer Science, 7:e644

    2021

  18. [26]

    Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. 2019. Albert: A lite bert for self-supervised learning of language representations. arXiv preprint arXiv:1909.11942

    2019 arXiv

  19. [27]

    Teven Le Scao, Angela Fan, Christopher Akiki, Ellie Pavlick, Suzana Ili \'c , Daniel Hesslow, Roman Castagn \'e , Alexandra Sasha Luccioni, Fran c ois Yvon, Matthias Gall \'e , et al. 2022. Bloom: A 176b-parameter open-access multilingual language model. arXiv preprint arXiv:2...

    2022 arXiv

  20. [28]

    Binny Mathew, Sandipan Sikdar, Florian Lemmerich, and Markus Strohmaier. 2020. The polar framework: Polar opposites enable interpretability of pre-trained word embeddings. In Proceedings of The Web Conference 2020 , pages 1548--1558

    2020

  21. [29]

    Bowman, and Rachel Rudinger

    Chandler May, Alex Wang, Shikha Bordia, Samuel R. Bowman, and Rachel Rudinger. 2019. https://doi.org/10.18653/v1/N19-1063 On measuring social biases in sentence encoders . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational...

    2019 doi

  22. [30]

    Miller, Richard Beckwith, Christiane Fellbaum, Derek Gross, and Katherine J

    George A. Miller, Richard Beckwith, Christiane Fellbaum, Derek Gross, and Katherine J. Miller. 1990. https://academic.oup.com/ijl/article-abstract/3/4/235/923280 Introduction to WordNet : An on-line lexical database . International journal of lexicography, 3(4):235--244

    1990

  23. [31]

    Gandalf Nicolas, Xuechunzi Bai, and Susan T. Fiske. 2021. https://doi.org/10.1002/ejsp.2724 Comprehensive stereotype content dictionaries using a semi‐automated method . European Journal of Social Psychology, 51(1):178--196

    2021 doi

  24. [32]

    Ali Omrani, Alireza Salkhordeh Ziabari, Charles Yu, Preni Golazizian, Brendan Kennedy, Mohammad Atari, Heng Ji, and Morteza Dehghani. 2023. https://doi.org/10.18653/v1/2023.acl-long.227 Social-group-agnostic bias mitigation via the stereotype content model . In Proceedings of ...

    2023 doi

  25. [33]

    Shiva Omrani Sabbaghi, Robert Wolfe, and Aylin Caliskan. 2023. Evaluating biased attitude associations of language models in an intersectional context. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, pages 542--553

    2023

  26. [34]

    Suci, and Percy H

    Charles Egerton Osgood, George J. Suci, and Percy H. Tannenbaum. 1957. The measurement of meaning. 47. University of Illinois press

    1957

  27. [35]

    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

  28. [36]

    Schuster and Simon Hegelich

    Carolin M. Schuster and Simon Hegelich. 2022. https://doi.org/10.18653/v1/2022.findings-acl.89 From BERT ` s P oint of V iew: R evealing the P revailing C ontextual D ifferences . In Findings of the Association for Computational Linguistics: ACL 2022, pages 1120--1138, Dublin,...

    2022 doi

  29. [37]

    Elisa Celis

    Yi Chern Tan and L. Elisa Celis. 2019. Assessing social and intersectional biases in contextualized word representations. Advances in neural information processing systems, 32

    2019

  30. [38]

    TeamGemma, Thomas Mesnard, Cassidy Hardin, Robert Dadashi, Surya Bhupatiraju, Shreya Pathak, Laurent Sifre, Morgane Rivi \`e re, Mihir Sanjay Kale, Juliette Love, et al. 2024 a . Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295

    2024 arXiv

  31. [39]

    TeamGemma, Morgane Riviere, Shreya Pathak, Pier Giuseppe Sessa, Cassidy Hardin, Surya Bhupatiraju, L \'e onard Hussenot, Thomas Mesnard, Bobak Shahriari, Alexandre Ram \'e , et al. 2024 b . Gemma 2: Improving open language models at a practical size. arXiv preprint arXiv:2408.00118

    2024 arXiv

  32. [40]

    Ian Tenney, Dipanjan Das, and Ellie Pavlick. 2019 a . https://www.aclweb.org/anthology/P19-1452.pdf BERT Rediscovers the Classical NLP Pipeline . In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4593--4601. Association for Compu...

    2019

  33. [41]

    Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Samuel R Bowman, Dipanjan Das, et al. 2019 b . What do you learn from context? probing for sentence structure in contextualized word representations. In International...

    2019

  34. [42]

    Eddie Ungless, Amy Rafferty, Hrichika Nag, and Bj \"o rn Ross. 2022. https://doi.org/10.18653/v1/2022.nlpcss-1.23 A robust bias mitigation procedure based on the stereotype content model . In Proceedings of the Fifth Workshop on Natural Language Processing and Computational So...

    2022 doi

  35. [43]

    Lütfi Kerem Şenel, Furkan Şahinuç, Veysel Yücesoy, Hinrich Schütze, Tolga Çukur, and Aykut Koç. 2022. Learning interpretable word embeddings via bidirectional alignment of dimensions with semantic concepts. Information Processing & Management, 59(3):102925

    2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.