Pith. sign in

REVIEW 3 major objections 1 cited by

Investigating Intersectional Bias in Large Language Models using Confidence Disparities in Coreference Resolution

T0 review · 3 major / 0 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Five LLMs show confidence gaps of up to 40 percentage points on coreference resolution for intersectional identities, with the greatest uncertainty for doubly disadvantaged groups in anti-stereotypical settings.

desk verdict The abstract promises an intersectional bias benchmark, but the full text is an unrelated robotics paper—there is no actual paper here to review. read the letter →

arxiv 2508.07111 v1 pith:ZBVOPFCS submitted 2025-08-09 cs.CL cs.AI

Falaah Arif Khan , Nivedha Sivakumar , Yinong Oliver Wang , Katherine Metcalf , Cezanne Camacho , Barry-John Theobald , Luca Zappella , Nicholas Apostoloff This is my paper
classification cs.CLcs.AI
keywords intersectionalbiascoreferenceresolutionconfidencedisparitylargelanguagemodelsWinoIdentitygroupfairnessmemorizationvsreasoninguncertainty
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 tries to establish that bias in large language models appears not only in what they say but in how confident they are, and that this confidence signal exposes intersectional disparities. The authors build WinoIdentity, a 245,700-prompt benchmark that crosses binary gender with 25 demographic markers across 10 attributes to form 50 distinct bias patterns. They introduce a metric, Coreference Confidence Disparity, that compares model confidence across intersectional identity groups. Evaluating five recent LLMs, they report disparities as high as 40 percentage points on attributes like body type, sexual orientation, and socio-economic status, with models most uncertain about doubly disadvantaged identities in anti-stereotypical settings. They argue these results show two independent failures: one in value alignment (bias) and one in validity (memorization rather than reasoning).

What carries the argument

Two components carry the argument. First, WinoIdentity, a benchmark built by augmenting WinoBias with 25 demographic markers across 10 attributes (including age, nationality, race, body type, sexual orientation, and socio-economic status) intersected with binary gender, yielding 245,700 prompts and 50 distinct bias patterns. Second, Coreference Confidence Disparity, a group (un)fairness metric that measures whether models are more or less confident for some intersectional identities than others. The paper's investigative lens is uncertainty: rather than looking only at task accuracy, it treats a model's output confidence as a signal of harm from omission and underrepresentation.

What would settle it

If the same confidence disparities appear when demographic markers are replaced by matched neutral tokens, or if the models are shown to be well calibrated within each identity group, the claim that confidence gaps reflect social bias would collapse; the memorization claim would fail if confidence remains high on novel identity-name combinations.

Watch

Extended reading notes

Core claim

The paper's central claim is that coreference resolution systems exhibit measurable, group-dependent confidence differences that signal social bias, and that these differences can be detected even when accuracy looks high. Evidence comes from a new benchmark, WinoIdentity, which augments WinoBias with 25 demographic markers across 10 attributes intersected with binary gender, producing 245,700 prompts covering 50 bias patterns. The proposed group unfairness metric, Coreference Confidence Disparity, quantifies whether models are more or less confident for some intersectional identities than others. The authors report confidence disparities of up to 40% across demographic attributes, with the

Load-bearing premise

The core claims stand on the premise that a model's reported confidence in coreference resolution is a faithful measure of its true uncertainty about an identity, and that lower confidence on privileged markers indicates memorization rather than reasoning.

Editorial extensions

If this is right

  • Fairness audits of LLMs in decision-support roles should measure confidence distributions, not just accuracy, because confidence gaps can persist even when outputs are correct.
  • WinoIdentity provides a reusable testbed for intersectional coreference bias across 50 identity combinations, allowing systematic comparisons across models and training regimes.
  • If lower confidence on privileged markers indeed reflects memorization rather than reasoning, reported accuracy gains in recent LLMs may overstate their reasoning ability.
  • Confidence disparities of up to 40% imply that in contexts like hiring or admissions, models could be systematically more certain about applicants from some intersectional groups than others, potentially affecting downstream decisions.
  • The benchmark's construction makes it possible to separate stereotype-consistent from anti-stereotypical settings, allowing targeted study of when uncertainty spikes.

Reading between the lines

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

  • Editorial: Confidence disparities may partly reflect calibration differences or task difficulty rather than social bias; a comparison that matches prompts for lexical frequency and syntactic complexity would help separate these explanations.
  • Editorial: The memorization inference is indirect; it would be strengthened by testing on identity-name combinations unseen in training or by probing whether confidence drops track training-data frequency.
  • Editorial: The intersectional extension could transfer to other NLP tasks with confidence signals, such as question answering or summarization, where omission and underrepresentation harms also matter.
  • Editorial: The benchmark's binary-gender intersection excludes non-binary and genderqueer identities; extending the attribute grid would likely reveal additional patterns of uncertainty.
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 / 0 minor

Summary. The submission, as identified by its abstract and arXiv metadata, claims a study of intersectional bias in large language models through confidence disparities in coreference resolution. The abstract states that the authors constructed a WinoIdentity benchmark with 25 demographic markers and 10 attributes, propose a Coreference Confidence Disparity metric, evaluate five LLMs, observe confidence disparities as high as 40%, and conclude that LLM performance may reflect memorization rather than logical reasoning. However, the full text provided is an entirely different paper: 'DexFruit: Dexterous Manipulation and Gaussian Splatting Inspection of Fruit,' a robotics manuscript by different authors concerning tactile sensing, grasping, and 3D reconstruction of fruit. The DexFruit text contains none of the components advertised in the abstract: no WinoIdentity benchmark, no coreference-confidence metric, no list of evaluated LLMs, no bias results, and no related analysis. The arXiv identifier in the file header, 2508.07118v3 [cs.RO], also differs from the claimed identifier 2508.07111 [cs.CL]. Because the manuscript body is unrelated to the central claim, no scientific assessment of the bias study is possible.

Significance. If the abstract's claims were supported, the work would address an important gap by extending single-axis bias evaluations to intersectional identity combinations and by proposing confidence disparity as a measurable bias signal. The reported 40% disparity and the distinction between value-alignment and validity failures could be significant contributions to the LLM-bias literature. However, as submitted, the manuscript contains no methods, experiments, benchmark description, metric definition, or results that substantiate these claims. The full text is a robotics paper with a different scope and authorship. There are no machine-checked proofs, reproducible code, parameter-free derivations, or falsifiable predictions to credit. The submission therefore offers no validatable scientific content for the claimed study, and any potential significance is entirely contingent on external evidence not present in the manuscript.

major comments (3)
  1. [Full Text (entire document vs. Abstract)] The central claim of the abstract—an empirical study of intersectional confidence disparities in five LLMs using the WinoIdentity benchmark—has no supporting methodology or results in the submitted full text. The full text is a robotics paper titled 'DexFruit: Dexterous Manipulation and Gaussian Splatting Inspection of Fruit' by a different author list. It describes grasping policies, tactile sensing, and 3D reconstruction, with no mention of coreference resolution, demographic attributes, confidence disparities, or bias. This is not a local presentation issue; it makes the central claim completely unverifiable from the manuscript.
  2. [Header/arXiv metadata] The manuscript header identifies the paper as arXiv:2508.07118v3 [cs.RO], while the submission is claimed to be arXiv:2508.07111 [cs.CL]. The abstract's empirical claims (245,700 prompts, 50 bias patterns, 40% disparity) cannot be traced to any section, equation, table, or dataset in the body. Since the document's apparent identity and content are inconsistent with the abstract, the submission does not constitute a coherent research paper that can be reviewed.
  3. [Abstract (unsupported central assertions)] Even granting the abstract's framing, two load-bearing premises are asserted without evidence: (i) that output confidence in coreference resolution is a valid proxy for model uncertainty and social bias, and (ii) that lower confidence for privileged markers indicates memorization rather than logical reasoning. Because the full text is unrelated, no derivation, experiment, or argument addresses these premises. As submitted, the paper's claims are unfalsifiable observations with no methodological grounding.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity can be established; the submitted full text does not contain the claimed methodology.

full rationale

The abstract describes a benchmark (WinoIdentity), a metric (Coreference Confidence Disparity), and empirical findings, but the full text supplied is an unrelated robotics paper (DexFruit) with no equations, no benchmark construction, no metric definition, and no results relevant to the abstract's claims. Circularity requires exhibiting a specific reduction of a claimed derivation to its own inputs or a self-citation chain; because none of the derivation steps are present, no such reduction can be quoted. The mismatch and resulting lack of support are serious validity/correctness concerns, but they are not evidence of circular reasoning. Accordingly, no circular step is identified and the score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 2 invented entities

Abstract-only audit enforced by the text mismatch. No free parameters are visible, but no fitted values can be excluded either, since the methodology is absent. The assumptions listed are the premises the abstract's conclusions depend on. The two invented artifacts (benchmark and metric) have no independent handle in the submitted material.

assumptions (3)
  • domain assumption Model confidence in coreference resolution is a valid proxy for model uncertainty and for socially meaningful bias.
    The abstract says the work investigates bias 'through the lens of uncertainty' and defines the Coreference Confidence Disparity metric on confidence values; the validity of this equation is not demonstrated in the text.
  • domain assumption Template prompts derived from WinoBias, augmented with demographic markers, transfer to real-world hiring and admissions contexts.
    The abstract motivates the work by hiring and admissions decisions but evaluates on synthetic coreference templates; the transfer is assumed without visible justification.
  • domain assumption Decreased confidence on privileged markers implies memorization rather than logical reasoning.
    The abstract's final sentences assert 'the recent impressive performance of LLMs is more likely due to memorization than logical reasoning' from observed confidence decreases; no supporting evidence appears in the text.
invented entities (2)
  • WinoIdentity benchmark
    purpose: Augmented WinoBias with 25 demographic markers across 10 attributes and binary gender, yielding 245,700 prompts for 50 bias patterns.
    Postulated in the abstract; no dataset artifact, URL, or release is present in the submitted text to verify it externally.
  • Coreference Confidence Disparity metric
    purpose: Group (un)fairness metric measuring whether models are more or less confident for some intersectional identities than others.
    Postulated in the abstract; no formal definition, formula, or validation appears in the submitted text.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Investigating Intersectional Bias in Large Language Models using Confidence Disparities in Coreference Resolution." pith.science (2026). https://pith.science/paper/ZBVOPFCS

@misc{pith2026250807111,
  author       = {Pith},
  title        = {Pith review of: Investigating Intersectional Bias in Large Language Models using Confidence Disparities in Coreference Resolution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZBVOPFCS}},
  note         = {Machine review of arXiv:2508.07111}
}
read the original abstract

Large language models (LLMs) have achieved impressive performance, leading to their widespread adoption as decision-support tools in resource-constrained contexts like hiring and admissions. There is, however, scientific consensus that AI systems can reflect and exacerbate societal biases, raising concerns about identity-based harm when used in critical social contexts. Prior work has laid a solid foundation for assessing bias in LLMs by evaluating demographic disparities in different language reasoning tasks. In this work, we extend single-axis fairness evaluations to examine intersectional bias, recognizing that when multiple axes of discrimination intersect, they create distinct patterns of disadvantage. We create a new benchmark called WinoIdentity by augmenting the WinoBias dataset with 25 demographic markers across 10 attributes, including age, nationality, and race, intersected with binary gender, yielding 245,700 prompts to evaluate 50 distinct bias patterns. Focusing on harms of omission due to underrepresentation, we investigate bias through the lens of uncertainty and propose a group (un)fairness metric called Coreference Confidence Disparity which measures whether models are more or less confident for some intersectional identities than others. We evaluate five recently published LLMs and find confidence disparities as high as 40% along various demographic attributes including body type, sexual orientation and socio-economic status, with models being most uncertain about doubly-disadvantaged identities in anti-stereotypical settings. Surprisingly, coreference confidence decreases even for hegemonic or privileged markers, indicating that the recent impressive performance of LLMs is more likely due to memorization than logical reasoning. Notably, these are two independent failures in value alignment and validity that can compound to cause social harm.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. A Novel Computational Thermodynamics Framework with Intrinsic Chemical Short-Range Order

    cond-mat.mtrl-sci 2025-08 unverdicted novelty 6.0 of 10

    A hybrid CVM-CALPHAD model using the Fowler-Yang-Li transform models chemical short-range order in multicomponent alloys at CVM accuracy and lower cost, benchmarked on fcc binaries and demonstrated on Cu-Au and Cu-Au-Ag.

Reference graph

Works this paper leans on

72 extracted references · 44 canonical work pages · cited by 1 Pith paper

  1. [1]

    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 format.date year duplicate empty "emp...

  2. [2]

    Junaid Ali, Preethi Lahoti, and Krishna P. Gummadi. Accounting for model uncertainty in algorithmic discrimination. In Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society, AIES '21, pp.\ 336–345, New York, NY, USA, 2021. Association for Computing Machinery. ISBN 9781450384735. doi:10.1145/3461702.3462630. URL https://doi.org/10.1145/346...

    arXiv 2021

  3. [3]

    Do large language models discriminate in hiring decisions on the basis of race, ethnicity, and gender? arXiv preprint arXiv:2406.10486, 2024

    Haozhe An, Christabel Acquaye, Colin Wang, Zongxia Li, and Rachel Rudinger. Do large language models discriminate in hiring decisions on the basis of race, ethnicity, and gender? arXiv preprint arXiv:2406.10486, 2024

    arXiv 2024

  4. [4]

    The Silicon Ceiling: Auditing GPT's Race and Gender Biases in Hiring

    Lena Armstrong, Abbey Liu, Stephen MacNeil, and Dana \"e Metaxa. The silicone ceiling: Auditing gpt's race and gender biases in hiring. arXiv preprint arXiv:2405.04412, 2024

    work page Pith review arXiv 2024

  5. [5]

    RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational Language Models

    Soumya Barikeri, Anne Lauscher, Ivan Vuli \'c , and Goran Glava s . Redditbias: A real-world resource for bias evaluation and debiasing of conversational language models. arXiv preprint arXiv:2106.03521, 2021

    work page Pith review arXiv 2021

  6. [6]

    Unmasking contextual stereotypes: Measuring and mitigating bert's gender bias

    Marion Bartl, Malvina Nissim, and Albert Gatt. Unmasking contextual stereotypes: Measuring and mitigating bert's gender bias. arXiv preprint arXiv:2010.14534, 2020

    arXiv 2010

  7. [7]

    On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pp.\ 610--623, 2021

    Emily M Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pp.\ 610--623, 2021

    2021

  8. [8]

    Are emily and greg more employable than lakisha and jamal? a field experiment on labor market discrimination

    Marianne Bertrand and Sendhil Mullainathan. Are emily and greg more employable than lakisha and jamal? a field experiment on labor market discrimination. American Economic Review, 94 0 (4): 0 991–1013, September 2004. doi:10.1257/0002828042002561. URL https://www.aeaweb.org/articles?id=10.1257/0002828042002561

    doi:10.1257/0002828042002561 2004

Show all 72 references
  1. [9]

    Language (technology) is power: A critical survey of `` bias '' in NLP

    Su Lin Blodgett, Solon Barocas, Hal Daum \'e III, and Hanna Wallach. Language (technology) is power: A critical survey of `` bias '' in NLP . In Dan Jurafsky, Joyce Chai, Natalie Schluter, and Joel Tetreault (eds.), ACL, 2020

    2020

  2. [10]

    Stereotyping norwegian salmon: An inventory of pitfalls in fairness benchmark datasets

    Su Lin Blodgett, Gilsinia Lopez, Alexandra Olteanu, Robert Sim, and Hanna Wallach. Stereotyping norwegian 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 In...

    2021

  3. [11]

    Language and identity

    Mary Bucholtz and Kira Hall. Language and identity. A companion to linguistic anthropology, 1: 0 369--394, 2004

    2004

  4. [12]

    Toward gender-inclusive coreference resolution: An analysis of gender and bias throughout the machine learning lifecycle

    Yang Trista Cao and Hal Daum \'e III. Toward gender-inclusive coreference resolution: An analysis of gender and bias throughout the machine learning lifecycle. Computational Linguistics, 47 0 (3): 0 615--661, 2021

    2021

  5. [13]

    Extracting intersectional stereotypes from embeddings: Developing and validating the flexible intersectional stereotype extraction procedure

    Tessa ES Charlesworth, Kshitish Ghate, Aylin Caliskan, and Mahzarin R Banaji. Extracting intersectional stereotypes from embeddings: Developing and validating the flexible intersectional stereotype extraction procedure. PNAS nexus, 3 0 (3): 0 pgae089, 2024

    2024

  6. [14]

    Intersectionality as critical social theory: Intersectionality as critical social theory, patricia hill collins, duke university press, 2019

    Patricia Hill Collins, Elaini Cristina Gonzaga da Silva, Emek Ergun, Inger Furseth, Kanisha D Bond, and Jone Mart \' nez-Palacios. Intersectionality as critical social theory: Intersectionality as critical social theory, patricia hill collins, duke university press, 2019. Cont...

    2019

  7. [15]

    A validity perspective on evaluating the justified use of data-driven decision-making algorithms

    Amanda Coston, Anna Kawakami, Haiyi Zhu, Ken Holstein, and Hoda Heidari. A validity perspective on evaluating the justified use of data-driven decision-making algorithms. In 2023 IEEE conference on secure and trustworthy machine learning (SaTML), pp.\ 690--704. IEEE, 2023

    2023

  8. [16]

    The algorithmic leviathan: Arbitrariness, fairness, and opportunity in algorithmic decision-making systems

    Kathleen Creel and Deborah Hellman. The algorithmic leviathan: Arbitrariness, fairness, and opportunity in algorithmic decision-making systems. Canadian Journal of Philosophy, 52 0 (1): 0 26--43, 2022

    2022

  9. [17]

    Demarginalizing the intersection of race and sex: A black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics

    Kimberl\'e Crenshaw. Demarginalizing the intersection of race and sex: A black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics. The University of Chicago Legal Forum, 140: 0 139--167, 1989

    1989

  10. [18]

    Are ai systems biased against the poor? a machine learning analysis using word2vec and glove embeddings

    Georgina Curto, Mario Fernando Jojoa Acosta, Flavio Comim, and Bego \ n a Garcia-Zapirain. Are ai systems biased against the poor? a machine learning analysis using word2vec and glove embeddings. AI & society, 39 0 (2): 0 617--632, 2024

    2024

  11. [19]

    Second order winobias (sowinobias) test set for latent gender bias detection in coreference resolution

    Hillary Dawkins. Second order winobias (sowinobias) test set for latent gender bias detection in coreference resolution. arXiv preprint arXiv:2109.14047, 2021

    2021 arXiv

  12. [20]

    Decomposition of uncertainty in bayesian deep learning for efficient and risk-sensitive learning

    Stefan Depeweg, Jose-Miguel Hernandez-Lobato, Finale Doshi-Velez, and Steffen Udluft. Decomposition of uncertainty in bayesian deep learning for efficient and risk-sensitive learning. In International conference on machine learning, pp.\ 1184--1193. PMLR, 2018

    2018

  13. [21]

    On measuring and mitigating biased inferences of word embeddings

    Sunipa Dev, Tao Li, Jeff M Phillips, and Vivek Srikumar. On measuring and mitigating biased inferences of word embeddings. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pp.\ 7659--7666, 2020

    2020

  14. [22]

    Fairness through awareness

    Cynthia Dwork, Moritz Hardt, Toniann Pitassi, Omer Reingold, and Richard Zemel. Fairness through awareness. In Proceedings of the 3rd innovations in theoretical computer science conference, pp.\ 214--226, 2012

    2012

  15. [23]

    Winoqueer: A community-in-the-loop benchmark for anti-lgbtq+ bias in large language models

    Virginia K Felkner, Ho-Chun Herbert Chang, Eugene Jang, and Jonathan May. Winoqueer: A community-in-the-loop benchmark for anti-lgbtq+ bias in large language models. arXiv preprint arXiv:2306.15087, 2023

    2023 arXiv

  16. [24]

    A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition

    Susan T Fiske, Amy JC Cuddy, Peter Glick, and Jun Xu. A model of (often mixed) stereotype content: Competence and warmth respectively follow from perceived status and competition. Journal of Personality and Social Psychology, 82 0 (6): 0 878--902, 2002

    2002

  17. [25]

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning

    Yarin Gal and Zoubin Ghahramani. Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In Maria Florina Balcan and Kilian Q. Weinberger (eds.), Proceedings of The 33rd International Conference on Machine Learning, volume 48 of Proceedings of Mac...

    2016

  18. [26]

    Counterfactual fairness in text classification through robustness

    Sahaj Garg, Vincent Perot, Nicole Limtiaco, Ankur Taly, Ed H Chi, and Alex Beutel. Counterfactual fairness in text classification through robustness. In Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society, pp.\ 219--226, 2019

    2019

  19. [27]

    A survey on intersectional fairness in machine learning: Notions, mitigation, and challenges

    Usman Gohar and Lu Cheng. A survey on intersectional fairness in machine learning: Notions, mitigation, and challenges. arXiv preprint arXiv:2305.06969, 2023

    2023 arXiv

  20. [28]

    Algorithmic arbitrariness in content moderation

    Juan Felipe Gomez, Caio Machado, Lucas Monteiro Paes, and Flavio Calmon. Algorithmic arbitrariness in content moderation. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, pp.\ 2234--2253, 2024

    2024

  21. [29]

    Akal badi ya bias: An exploratory study of gender bias in hindi language technology

    Rishav Hada, Safiya Husain, Varun Gumma, Harshita Diddee, Aditya Yadavalli, Agrima Seth, Nidhi Kulkarni, Ujwal Gadiraju, Aditya Vashistha, Vivek Seshadri, et al. Akal badi ya bias: An exploratory study of gender bias in hindi language technology. In The 2024 ACM Conference on ...

    2024

  22. [30]

    Equality of opportunity in supervised learning

    Moritz Hardt, Eric Price, and Nati Srebro. Equality of opportunity in supervised learning. Advances in neural information processing systems, 29, 2016

    2016

  23. [31]

    Misgendered: Limits of large language models in understanding pronouns

    Tamanna Hossain, Sunipa Dev, and Sameer Singh. Misgendered: Limits of large language models in understanding pronouns. arXiv preprint arXiv:2306.03950, 2023

    2023 arXiv

  24. [32]

    Socialcounterfactuals: Probing and mitigating intersectional social biases in vision-language models with counterfactual examples

    Phillip Howard, Avinash Madasu, Tiep Le, Gustavo Lujan Moreno, Anahita Bhiwandiwalla, and Vasudev Lal. Socialcounterfactuals: Probing and mitigating intersectional social biases in vision-language models with counterfactual examples. In Proceedings of the IEEE/CVF Conference o...

    2024

  25. [33]

    Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods

    Eyke H \"u llermeier and Willem Waegeman. Aleatoric and epistemic uncertainty in machine learning: An introduction to concepts and methods. Machine learning, 110 0 (3): 0 457--506, 2021

    2021

  26. [34]

    Are female carpenters like blue bananas? a corpus investigation of occupation gender typicality

    Da Ju, Karen Ullrich, and Adina Williams. Are female carpenters like blue bananas? a corpus investigation of occupation gender typicality. In Findings of the Association for Computational Linguistics ACL 2024, pp.\ 4254--4274, 2024

    2024

  27. [35]

    Taxonomizing and measuring representational harms: A look at image tagging

    Jared Katzman, Angelina Wang, Morgan Scheuerman, Su Lin Blodgett, Kristen Laird, Hanna Wallach, and Solon Barocas. Taxonomizing and measuring representational harms: A look at image tagging. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp.\ 1427...

    2023

  28. [36]

    What uncertainties do we need in bayesian deep learning for computer vision? In I

    Alex Kendall and Yarin Gal. What uncertainties do we need in bayesian deep learning for computer vision? In I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (eds.), Advances in Neural Information Processing Systems, volume 30. Curran ...

    2017

  29. [37]

    Examining gender and race bias in two hundred sentiment analysis systems

    Svetlana Kiritchenko and Saif M Mohammad. Examining gender and race bias in two hundred sentiment analysis systems. arXiv preprint arXiv:1805.04508, 2018

    2018 arXiv

  30. [38]

    Dreyer, Aleksandar Shtedritski, and Yuki M

    Hannah Rose Kirk, Yennie Jun, Haider Iqbal, Elias Benussi, Filippo Volpin, Frederic A. Dreyer, Aleksandar Shtedritski, and Yuki M. Asano. Bias out-of-the-box: an empirical analysis of intersectional occupational biases in popular generative language models. In Proceedings of t...

    2024

  31. [39]

    Stereotype content at the intersection of gender and sexual orientation

    Amanda Klysing, Anna Lindqvist, and Fredrik Bj \"o rklund. Stereotype content at the intersection of gender and sexual orientation. Frontiers in Psychology, 12: 0 713839, 2021

    2021

  32. [40]

    intersectionally fair

    Youjin Kong. Are “intersectionally fair” ai algorithms really fair to women of color? a philosophical analysis. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, pp.\ 485--494, 2022

    2022

  33. [41]

    Gender bias and stereotypes in large language models

    Hadas Kotek, Rikker Dockum, and David Sun. Gender bias and stereotypes in large language models. In Proceedings of the ACM collective intelligence conference, pp.\ 12--24, 2023

    2023

  34. [42]

    Uncertainty as a fairness measure

    Selim Kuzucu, Jiaee Cheong, Hatice Gunes, and Sinan Kalkan. Uncertainty as a fairness measure. Journal of Artificial Intelligence Research, 81: 0 307--335, 2024

    2024

  35. [43]

    Simple and scalable predictive uncertainty estimation using deep ensembles

    Balaji Lakshminarayanan, Alexander Pritzel, and Charles Blundell. Simple and scalable predictive uncertainty estimation using deep ensembles. Advances in neural information processing systems, 30, 2017

    2017

  36. [44]

    Collecting a large-scale gender bias dataset for coreference resolution and machine translation

    Shahar Levy, Koren Lazar, and Gabriel Stanovsky. Collecting a large-scale gender bias dataset for coreference resolution and machine translation. arXiv preprint arXiv:2109.03858, 2021

    2021 arXiv

  37. [45]

    A survey on fairness in large language models

    Yingji Li, Mengnan Du, Rui Song, Xin Wang, and Ying Wang. A survey on fairness in large language models. arXiv preprint arXiv:2308.10149, 2023

    2023 arXiv

  38. [46]

    Comparing diversity, negativity, and stereotypes in chinese-language ai technologies: a case study on baidu, ernie and qwen

    Geng Liu, Carlo Alberto Bono, and Francesco Pierri. Comparing diversity, negativity, and stereotypes in chinese-language ai technologies: a case study on baidu, ernie and qwen. arXiv preprint arXiv:2408.15696, 2024

    2024 arXiv

  39. [47]

    Intersectional stereotypes in large language models: Dataset and analysis

    Weicheng Ma, Brian Chiang, Tong Wu, Lili Wang, and Soroush Vosoughi. Intersectional stereotypes in large language models: Dataset and analysis. In Findings of the Association for Computational Linguistics: EMNLP 2023, pp.\ 8589--8597, 2023

    2023

  40. [48]

    Gsm-symbolic: Understanding the limitations of mathematical reasoning in large language models

    Iman Mirzadeh, Keivan Alizadeh, Hooman Shahrokhi, Oncel Tuzel, Samy Bengio, and Mehrdad Farajtabar. Gsm-symbolic: Understanding the limitations of mathematical reasoning in large language models. arXiv preprint arXiv:2410.05229, 2024

    2024 arXiv

  41. [49]

    Torr, and Yarin Gal

    Jishnu Mukhoti, Andreas Kirsch, Joost van Amersfoort, Philip H.S. Torr, and Yarin Gal. Deep deterministic uncertainty: A new simple baseline. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 24384--24394, June 2023

    2023

  42. [50]

    Stereoset: Measuring stereotypical bias in pretrained language models

    Moin Nadeem, Anna Bethke, and Siva Reddy. Stereoset: Measuring stereotypical bias in pretrained language models. arXiv preprint arXiv:2004.09456, 2020

    2004 arXiv

  43. [51]

    Crows-pairs: A challenge dataset for measuring social biases in masked language models

    Nikita Nangia, Clara Vania, Rasika Bhalerao, and Samuel R Bowman. Crows-pairs: A challenge dataset for measuring social biases in masked language models. arXiv preprint arXiv:2010.00133, 2020

    2010 arXiv

  44. [52]

    Factoring the matrix of domination: A critical review and reimagination of intersectionality in ai fairness

    Anaelia Ovalle, Arjun Subramonian, Vagrant Gautam, Gilbert Gee, and Kai-Wei Chang. Factoring the matrix of domination: A critical review and reimagination of intersectionality in ai fairness. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, pp.\ 496--...

    2023

  45. [53]

    BBQ : A hand-built bias benchmark for question answering

    Alicia Parrish, Angelica Chen, Nikita Nangia, Vishakh Padmakumar, Jason Phang, Jana Thompson, Phu Mon Htut, and Samuel Bowman. BBQ : A hand-built bias benchmark for question answering. In Smaranda Muresan, Preslav Nakov, and Aline Villavicencio (eds.), Findings of the Associat...

    2022 doi

  46. [54]

    Perturbation augmentation for fairer nlp

    Rebecca Qian, Candace Ross, Jude Fernandes, Eric Smith, Douwe Kiela, and Adina Williams. Perturbation augmentation for fairer nlp. arXiv preprint arXiv:2205.12586, 2022

    2022 arXiv

  47. [55]

    Gender bias in coreference resolution

    Rachel Rudinger, Jason Naradowsky, Brian Leonard, and Benjamin Van Durme. Gender bias in coreference resolution. arXiv preprint arXiv:1804.09301, 2018

    2018 arXiv

  48. [56]

    The unequal opportunities of large language models: Examining demographic biases in job recommendations by chatgpt and llama

    Abel Salinas, Parth Shah, Yuzhong Huang, Robert McCormack, and Fred Morstatter. The unequal opportunities of large language models: Examining demographic biases in job recommendations by chatgpt and llama. In Proceedings of the 3rd ACM Conference on Equity and Access in Algori...

    2023

  49. [57]

    Are emergent abilities of large language models a mirage? Advances in Neural Information Processing Systems, 36: 0 55565--55581, 2023

    Rylan Schaeffer, Brando Miranda, and Sanmi Koyejo. Are emergent abilities of large language models a mirage? Advances in Neural Information Processing Systems, 36: 0 55565--55581, 2023

    2023

  50. [58]

    The woman worked as a babysitter: On biases in language generation

    Emily Sheng, Kai-Wei Chang, Premkumar Natarajan, and Nanyun Peng. The woman worked as a babysitter: On biases in language generation. arXiv preprint arXiv:1909.01326, 2019

    1909 arXiv

  51. [59]

    A framework for understanding sources of harm throughout the machine learning life cycle

    Harini Suresh and John Guttag. A framework for understanding sources of harm throughout the machine learning life cycle. In Proceedings of the 1st ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization, pp.\ 1--9, 2021

    2021

  52. [60]

    Fairness through aleatoric uncertainty

    Anique Tahir, Lu Cheng, and Huan Liu. Fairness through aleatoric uncertainty. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management, CIKM '23, pp.\ 2372–2381, New York, NY, USA, 2023. Association for Computing Machinery. ISBN 979840070...

    2023

  53. [61]

    Neutral rewriter: A rule-based and neural approach to automatic rewriting into gender-neutral alternatives

    Eva Vanmassenhove, Chris Emmery, and Dimitar Shterionov. Neutral rewriter: A rule-based and neural approach to automatic rewriting into gender-neutral alternatives. arXiv preprint arXiv:2109.06105, 2021

    2021 arXiv

  54. [62]

    Measuring representational harms in image captioning

    Angelina Wang, Solon Barocas, Kristen Laird, and Hanna Wallach. Measuring representational harms in image captioning. In Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, pp.\ 324--335, 2022

    2022

  55. [63]

    Aleatoric and epistemic discrimination: Fundamental limits of fairness interventions

    Hao Wang, Luxi He, Rui Gao, and Flavio Calmon. Aleatoric and epistemic discrimination: Fundamental limits of fairness interventions. Advances in Neural Information Processing Systems, 36, 2024

    2024

  56. [64]

    Mind the gap: A balanced corpus of gendered ambiguous pronouns

    Kellie Webster, Marta Recasens, Vera Axelrod, and Jason Baldridge. Mind the gap: A balanced corpus of gendered ambiguous pronouns. Transactions of the Association for Computational Linguistics, 6: 0 605--617, 2018

    2018

  57. [65]

    Easy problems that llms get wrong

    Sean Williams and James Huckle. Easy problems that llms get wrong. arXiv preprint arXiv:2405.19616, 2024

    2024 arXiv

  58. [66]

    Gender, race, and intersectional bias in resume screening via language model retrieval

    Kyra Wilson and Aylin Caliskan. Gender, race, and intersectional bias in resume screening via language model retrieval. In Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, volume 7, pp.\ 1578--1590, 2024

    2024

  59. [67]

    What is your favorite gender, mlm? gender bias evaluation in multilingual masked language models

    Jeongrok Yu, Seong Ug Kim, Jacob Choi, and Jinho D Choi. What is your favorite gender, mlm? gender bias evaluation in multilingual masked language models. Information, 15 0 (9): 0 549, 2024

    2024

  60. [68]

    Fairness beyond disparate treatment & disparate impact: Learning classification without disparate mistreatment

    Muhammad Bilal Zafar, Isabel Valera, Manuel Gomez Rodriguez, and Krishna P Gummadi. Fairness beyond disparate treatment & disparate impact: Learning classification without disparate mistreatment. In Proceedings of the 26th international conference on world wide web, pp.\ 1171-...

    2017

  61. [69]

    Gender bias in coreference resolution: Evaluation and debiasing methods

    Jieyu Zhao, Tianlu Wang, Mark Yatskar, Vicente Ordonez, and Kai-Wei Chang. Gender bias in coreference resolution: Evaluation and debiasing methods. In Marilyn Walker, Heng Ji, and Amanda Stent (eds.), Proceedings of the 2018 Conference of the North A merican Chapter of the Ass...

    2018 doi

  62. [70]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  63. [71]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  64. [72]

    [pronoun]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

    2022

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.