Pith. sign in

REVIEW 3 major objections 6 minor 52 references

Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A new human-annotated Chinese corpus, CORGI-PM, grounds three gender-bias benchmarks and shows that rewriting biased sentences is far harder than detecting them.

desk verdict A useful shared-task leaderboard and a real Chinese bias resource, but the paper's own numbers do not add up and the annotation-quality evidence is weaker than it looks. read the letter →

arxiv 2506.12574 v1 pith:2I52VSHM submitted 2025-06-14 cs.CL

Yizhi Li , Ge Zhang , Hanhua Hong , Yiwen Wang , Chenghua Lin This is my paper
classification cs.CL
keywords genderbiasChinesecorpusmitigationdetectionclassificationhumanannotationsharedtaskstereotypes
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 introduces CORGI-PM, a human-annotated Chinese corpus of 32.9k sentences built for studying textual gender bias. About 5.2k of those sentences carry fine-grained bias labels and human-written bias-free rewrites, which turns the corpus into a resource for mitigation as well as detection. The authors define three shared tasks—detection, classification, and mitigation—and report results from participating teams. The results show a clear difficulty progression: detection is tractable, classification is harder, and mitigation is the hardest by a wide margin.

What carries the argument

The central object is CORGI-PM, a corpus of 32.9k Chinese sentences, and the mechanism that produces it is a two-stage filtering pipeline followed by a structured human annotation scheme. Word embeddings from Chinese language models are used to compute gender-bias scores by comparing representations against a 'she' minus 'he' direction, recalling candidate sentences; sentence-level reranking then selects the final pool. Six Chinese-speaking annotators label each sentence as biased or not, assign one or more of three stereotype subclasses—gender-stereotyped activity and career choices, descriptions and inductions, or expressed attitudes and beliefs—and write a neutral paraphrase for biased sentences. The inclusion of these human-written debiased versions is what makes mitigation a measurable task rather than just an aspiration.

What would settle it

If an independent re-annotation of a random subset of the 5.2k biased sentences, using annotators not trained with the same scheme, produces materially different biased/non-biased decisions or different rewrites, then the claimed reliability of the corpus would be called into question.

Watch

Extended reading notes

Core claim

The paper's central claim is that a carefully annotated Chinese corpus, created by filtering large-scale corpora with gender-bias scoring and then having six native-speaker annotators classify and rewrite sentences, can support three concrete benchmarks for gender-bias work. The corpus contains 21.4k non-biased training sentences and 5.2k biased sentences, each labeled with one or more of three stereotype subclasses and paired with a human-written bias-eliminated version. The shared-task results show a clear difficulty gradient, with the best reported detection F1 near 0.85, classification near 0.65, and mitigation near 0.29, supporting the claim that mitigation is the bottleneck and requires more than classification ability.

Load-bearing premise

The benchmark's ground truth is only as good as six annotators' agreement after throwing away every sentence they were unsure about or disagreed on.

Editorial extensions

If this is right

  • Chinese NLP gains a reusable fairness benchmark with three difficulty-graded tasks and human-written reference corrections.
  • Models can be compared on a clear progression from detection to classification to mitigation, with mitigation identified as the bottleneck.
  • The parallel biased/unbiased sentence pairs provide direct supervision for training generation models to neutralize gender stereotypes while preserving meaning.
  • Fine-grained subclass labels allow evaluation of whether a system understands how bias is expressed, not only whether bias is present.
  • The balanced train/valid/test splits and additional held-out samples give a stable protocol for future leaderboard comparisons.

Reading between the lines

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

  • An implication the paper leaves implicit is that its quality-control step—removing all uncertain and contradictory annotations—may trim away the hardest cases, so the corpus could understate genuine annotator disagreement on subtle bias.
  • Because candidate sentences were selected using a 'she' minus 'he' embedding direction, biases not expressed through binary gendered words may be underrepresented, and the benchmark may not transfer to non-binary or purely contextual bias.
  • A natural next experiment is to train a debiasing model on the parallel biased-to-neutral pairs and measure whether downstream Chinese NLP tasks, such as coreference or sentiment analysis, become fairer without losing accuracy.
  • The large gap between detection and mitigation scores hints that reference-based automatic metrics may reward conservative edits; adopting the human evaluation as an official metric could change the leaderboard ranking.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. This paper presents CORGI-PM, a Chinese corpus for gender bias probing and mitigation, and reports on the NLPCC 2025 shared task built around it. The corpus is claimed to contain 32.9k human-annotated sentences, including 5.2k gender-biased sentences with parallel human-written debiased versions. The authors define three tasks—gender bias detection, classification, and mitigation—and provide baseline experiments plus a leaderboard of participating teams. The central claims are that the corpus is high-quality, that the three-task framework has a meaningful difficulty progression, and that detection is much easier than mitigation.

Significance. If the corpus and benchmarks are valid, this is a useful contribution to Chinese fairness NLP: it is one of the few human-annotated Chinese gender-bias resources, provides parallel debiased sentences, and includes both probing and mitigation tasks with a public leaderboard. The paper also ships explicit limitations and reports a concrete set of baseline and participant results, which is helpful for reproducibility. However, the value of the contribution hinges on the correctness of the reported dataset statistics and on the reliability of the annotation procedure; both currently have unresolved issues that a reader needs clarified before trusting the benchmark conclusions.

major comments (3)
  1. [Abstract and Table 1] The corpus statistics are internally inconsistent and do not reconcile with the abstract. In Table 1, the train split rows sum to 28.47k (1.90k + 2.70k + 2.47k + 21.4k), yet the Overall row reports 30.1k for Train. Adding Valid (1,391) and Test (1,409) gives 31.27k total, while the abstract states 32.9k. Moreover, the abstract and Section 1 state that the corpus contains 5.2k biased sentences, but the Table 1 rows for biased categories sum to 8.83k (7.07k train + 1.76k valid/test). The footnote about 100 additional samples per task does not close these gaps. These inconsistencies are load-bearing because the paper's central claim is that CORGI-PM is a large, high-quality corpus with a specific biased/non-biased composition.
  2. [Section 3.2, Quality Monitoring and Control, and Table 3] The reported inter-annotator agreement (IAA) is not a trustworthy measure of annotation reliability because it is computed after excluding uncertain and contradictory samples. Item (3) states that annotators were encouraged not to enter uncertain samples and that no contradictory samples were entered; Table 3 then reports IAA values of 0.802, 0.935, and 0.987. The coefficient is unnamed (simple agreement, Cohen's kappa, and Fleiss' kappa can differ substantially), and the text says 'among three annotators' while item (1) says there were six annotators. Because the excluded cases are precisely the ambiguous ones where bias judgment is hardest, the IAA figures overstate raw agreement, and the benchmark's validity depends on the unverified assumption that discarded disagreements carry no signal.
  3. [Section 5.2, Challenge of Mitigation] The mitigation task is evaluated with BLEU, ROUGE-L, and METEOR against a single human reference, but Section 3.2 explicitly instructs annotators to diversify expressions when writing debiased versions. Valid debiased outputs that differ lexically from the reference can therefore receive low scores even if they are semantically equivalent and unbiased. This means the observed low mitigation scores (e.g., 0.294 for the best team) and the claimed difficulty ordering (detection F1 ~0.85 versus mitigation ~0.29) may reflect the metric's sensitivity to lexical overlap rather than the intrinsic difficulty of mitigation. The paper should either use multiple references, report human evaluation results for the leaderboard, or temper the difficulty-progression conclusion.
minor comments (6)
  1. [Abstract] The phrase 'In the literature, we present the results' should be 'In this paper, we present the results'.
  2. [Section 1, last paragraph] The sentence 'all the 5.2k biased sentences in our corpus are further labeled with gender bias subclasses and companies with parallel bias-free versions' uses 'companies' where 'accompanied' is intended.
  3. [Section 5.1, Experiment Settings] The sentence 'We test the performance on both "yes" and "no" detection' is unclear; it should specify that per-class metrics for the biased and non-biased classes are reported.
  4. [Footnote 4] The footnote states that 100 additional samples per task are provided in addition to the original splits [42], but it is not explained how these samples relate to the counts in Table 1; please clarify whether they are included in the table or are separate.
  5. [References] References [37] and [38] are the same paper (Webster et al., Mind the GAP); one should be removed.
  6. [Table 2] The row labeled 'Sent. Length' reports single numbers for each column without specifying whether these are means, medians, or maxima; please clarify the statistic and its unit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: corpus is human-annotated, benchmarks are external, and the only self-citation is provenance for the dataset itself.

full rationale

The paper's central contributions are empirical: a Chinese gender-bias corpus built by automatic candidate retrieval followed by human annotation, and three benchmark tasks evaluated on held-out test splits by external submitted systems. There is no derivation chain in which a predicted quantity is equivalent by construction to a fitted input. The candidate-pool method borrows word- and sentence-level bias scoring from external work [4,14], not from the authors' own fitted values. The detection, classification, and mitigation leaderboards measure held-out model outputs, so those scores are not forced by the corpus construction. The one self-citation, [42], is the authors' earlier CORGI-PM paper; it is used to attribute the original data splits and to describe the 100 additional contamination-control samples. This is provenance and attribution rather than a load-bearing argument: the corpus itself is described, with statistics and annotation examples, in the present paper, and its quality is an empirical claim that can be checked against the released data. The annotation-protocol concern noted by the skeptic (that uncertain and contradictory samples were excluded before inter-annotator agreement was computed) is a validity and selection-bias concern, not a circularity: excluding disagreement does not make any reported benchmark score equal to the annotation input by construction. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known result is merely re-named in new coordinates. The paper is therefore self-contained against external benchmarks, and the most honest finding is no significant circularity.

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

The paper does not derive numbers from first principles. Its corpus depends on two unreported filtering thresholds and on the assumption that embedding-based gender directions and the three-category taxonomy correctly capture Chinese gender bias. No new theoretical entities are introduced.

free parameters (2)
  • global threshold score = not reported
    Used in Section 3.1 to filter candidate sentences for annotation; the exact value is not given, so the composition of the corpus depends on an unreported hand-chosen cutoff.
  • in-group threshold rank = not reported
    Used in Section 3.1 after grouping by bias keyword; this rank cutoff determines which sentences are retained and is not reported.
assumptions (3)
  • domain assumption The 'she' minus 'he' embedding direction is a valid measure of gender bias in Chinese.
    The candidate pool in Section 3.1 is built by ranking words and sentences against this seed direction from references [4,14]; if the direction captures generic gender association rather than bias, the corpus will be skewed.
  • domain assumption The three categories AC, DI, and ANB form an adequate taxonomy for Chinese gender bias.
    All biased sentences are labeled with these categories in Section 3.2; coverage and mutual exclusivity are not empirically validated.
  • domain assumption Human annotators' labels are the ground truth for gender bias.
    The corpus labels and rewrites are produced by six annotators in Section 3.2 without external validation; the paper discards contradictory cases, so the final labels represent a selected consensus.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge." pith.science (2026). https://pith.science/paper/2I52VSHM

@misc{pith2026250612574,
  author       = {Pith},
  title        = {Pith review of: Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2I52VSHM}},
  note         = {Machine review of arXiv:2506.12574}
}
read the original abstract

As natural language processing for gender bias becomes a significant interdisciplinary topic, the prevalent data-driven techniques, such as pre-trained language models, suffer from biased corpus. This case becomes more obvious regarding those languages with less fairness-related computational linguistic resources, such as Chinese. To this end, we propose a Chinese cOrpus foR Gender bIas Probing and Mitigation (CORGI-PM), which contains 32.9k sentences with high-quality labels derived by following an annotation scheme specifically developed for gender bias in the Chinese context. It is worth noting that CORGI-PM contains 5.2k gender-biased sentences along with the corresponding bias-eliminated versions rewritten by human annotators. We pose three challenges as a shared task to automate the mitigation of textual gender bias, which requires the models to detect, classify, and mitigate textual gender bias. In the literature, we present the results and analysis for the teams participating this shared task in NLPCC 2025.

Figures

Figures reproduced from arXiv: 2506.12574 by the authors.

Figure 1
Figure 1. Pipeline of Retrieving and Filtering Potentially Biased Sentences Candidate Pool [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Case Study of Mitigation Annotation Patterns. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 26 canonical work pages

  1. [42]

    Zhang, G., Li, Y ., Wu, Y ., Zhang, L., Lin, C., Geng, J., Wang, S., Fu, J.: Corgi-pm: A chinese corpus for gender bias probing and mitigation (2023), https://arxiv.org/abs/2301.00395

    arXiv 2023

  2. [1]

    Proceedings of WMT-08 (2007) 10 Y

    Agarwal, A., Lavie, A.: Meteor: An automatic metric for mt evaluation with high levels of correlation with human judgments. Proceedings of WMT-08 (2007) 10 Y . Li et al

    work page 2007

  3. [2]

    arXiv preprint arXiv:2309.16609 (2023)

    Bai, J., Bai, S., Chu, Y ., Cui, Z., Dang, K., Deng, X., Fan, Y ., Ge, W., Han, Y ., Huang, F., et al.: Qwen technical report. arXiv preprint arXiv:2309.16609 (2023)

    arXiv 2023

  4. [3]

    In: ACL (June 2020), https://www.microsoft.com/en-us/ research/publication/language-technology-is-power-a-critical-survey-of-bias-in-nlp/

    Blodgett, S.L., Barocas, S., Daumé III, H., Wallach, H.: Language (technology) is power: A critical survey of “bias” in nlp. In: ACL (June 2020), https://www.microsoft.com/en-us/ research/publication/language-technology-is-power-a-critical-survey-of-bias-in-nlp/

    work page 2020

  5. [4]

    In: NIPS (2016)

    Bolukbasi, T., Chang, K.W., Zou, J.Y ., Saligrama, V ., Kalai, A.T.: Man is to computer programmer as woman is to homemaker? debiasing word embeddings. In: NIPS (2016)

    work page 2016

  6. [5]

    be nice to your wife! the restaurants are closed

    Chiril, P., Benamara, F., Moriceau, V .: “be nice to your wife! the restaurants are closed”: Can gender stereotype detection improve sexism classification? In: Findings of the Association for Computational Linguistics: EMNLP 2021. pp. 2833–2844 (2021)

    work page 2021

  7. [6]

    In: Proceedings of the 12th language resources and evaluation conference

    Chiril, P., Moriceau, V ., Benamara, F., Mari, A., Origgi, G., Coulomb-Gully, M.: An annotated corpus for sexism detection in french tweets. In: Proceedings of the 12th language resources and evaluation conference. pp. 1397–1403 (2020)

    work page 2020

  8. [7]

    Nature Machine Intelligence 1(11), 495–496 (2019)

    Costa-jussà, M.R.: An analysis of gender bias studies in natural language processing. Nature Machine Intelligence 1(11), 495–496 (2019)

    work page 2019

Show all 52 references
  1. [8]

    arXiv preprint arXiv:2004.13922 (2020)

    Cui, Y ., Che, W., Liu, T., Qin, B., Wang, S., Hu, G.: Revisiting pre-trained models for chinese natural language processing. arXiv preprint arXiv:2004.13922 (2020)

    2020 arXiv

  2. [9]

    In: Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society

    Dixon, L., Li, J., Sorensen, J., Thain, N., Vasserman, L.: Measuring and mitigating unintended bias in text classification. In: Proceedings of the 2018 AAAI/ACM Conference on AI, Ethics, and Society. p. 67–73. AIES ’18, Association for Computing Machinery, New York, NY , USA (...

    2018

  3. [10]

    arXiv preprint arXiv:2402.11190 (2024)

    Dong, X., Wang, Y ., Yu, P.S., Caverlee, J.: Disclosure and mitigation of gender bias in llms. arXiv preprint arXiv:2402.11190 (2024)

    2024 arXiv

  4. [11]

    arXiv preprint arXiv:2502.14739 (2025)

    Du, X., Yao, Y ., Ma, K., Wang, B., Zheng, T., Zhu, K., Liu, M., Liang, Y ., Jin, X., Wei, Z., et al.: Supergpqa: Scaling llm evaluation across 285 graduate disciplines. arXiv preprint arXiv:2502.14739 (2025)

    2025 arXiv

  5. [12]

    arXiv preprint arXiv:2101.00027 (2020)

    Gao, L., Biderman, S., Black, S., Golding, L., Hoppe, T., Foster, C., Phang, J., He, H., Thite, A., Nabeshima, N., et al.: The pile: An 800gb dataset of diverse text for language modeling. arXiv preprint arXiv:2101.00027 (2020)

    2020 arXiv

  6. [13]

    Online Social Networks and Media 27, 100182 (2022)

    Jiang, A., Yang, X., Liu, Y ., Zubiaga, A.: Swsr: A chinese dataset and lexicon for online sexism detection. Online Social Networks and Media 27, 100182 (2022)

    2022

  7. [14]

    In: Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing

    Jiao, M., Luo, Z.: Gender bias hidden behind chinese word embeddings: The case of chi- nese adjectives. In: Proceedings of the 3rd Workshop on Gender Bias in Natural Language Processing. pp. 8–15 (2021)

    2021

  8. [15]

    Australasian Journal of Early Childhood 46(2), 112–125 (2021)

    King, T.L., Scovelle, A.J., Meehl, A., Milner, A.J., Priest, N.: Gender stereotypes and biases in early childhood: A systematic review. Australasian Journal of Early Childhood 46(2), 112–125 (2021)

    2021

  9. [16]

    In: Proceedings of the Seventh Joint Conference on Lexical and Computational Semantics

    Kiritchenko, S., Mohammad, S.: Examining gender and race bias in two hundred senti- ment analysis systems. In: Proceedings of the Seventh Joint Conference on Lexical and Computational Semantics. pp. 43–53. Association for Computational Linguistics, New Or- leans, Louisiana (Ju...

    2018 doi

  10. [17]

    In: Proceedings of the ACM collective intelligence conference

    Kotek, H., Dockum, R., Sun, D.: Gender bias and stereotypes in large language models. In: Proceedings of the ACM collective intelligence conference. pp. 12–24 (2023)

    2023

  11. [18]

    In: He, Y ., Ji, H., Li, S., Liu, Y ., Chang, C.H

    Li, Y ., Zhang, G., Yang, B., Lin, C., Ragni, A., Wang, S., Fu, J.: HERB: Measuring hierarchical regional bias in pre-trained language models. In: He, Y ., Ji, H., Li, S., Liu, Y ., Chang, C.H. (eds.) Findings of the Association for Computational Linguistics: AACL-IJCNLP 2022....

    2022 doi

  12. [19]

    In: Text summarization branches out

    Lin, C.Y .: Rouge: A package for automatic evaluation of summaries. In: Text summarization branches out. pp. 74–81 (2004) Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge 11

    2004

  13. [20]

    In: Logic, Language, and Security, pp

    Lu, K., Mardziel, P., Wu, F., Amancharla, P., Datta, A.: Gender bias in neural natural language processing. In: Logic, Language, and Security, pp. 189–202. Springer (2020)

    2020

  14. [21]

    In: Friedler, S.A., Wilson, C

    Madaan, N., Mehta, S., Agrawaal, T., Malhotra, V ., Aggarwal, A., Gupta, Y ., Saxena, M.: Analyze, detect and remove gender stereotyping from bollywood movies. In: Friedler, S.A., Wilson, C. (eds.) Proceedings of the 1st Conference on Fairness, Accountability and Trans- parenc...

    2018

  15. [22]

    In: Proceedings of the 40th annual meeting of the Association for Computational Linguistics

    Papineni, K., Roukos, S., Ward, T., Zhu, W.J.: Bleu: a method for automatic evaluation of machine translation. In: Proceedings of the 40th annual meeting of the Association for Computational Linguistics. pp. 311–318 (2002)

    2002

  16. [23]

    arXiv preprint arXiv:1910.04602 (2019)

    Parikh, P., Abburi, H., Badjatiya, P., Krishnan, R., Chhaya, N., Gupta, M., Varma, V .: Multi-label categorization of accounts of sexism using a neural framework. arXiv preprint arXiv:1910.04602 (2019)

    2019 arXiv

  17. [24]

    In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)

    Parikh, P., Abburi, H., Badjatiya, P., Krishnan, R., Chhaya, N., Gupta, M., Varma, V .: Multi- label categorization of accounts of sexism using a neural framework. In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th Internatio...

    2019 doi

  18. [25]

    In: Pro- ceedings of the 2018 Conference on Empirical Methods in Natural Language Processing

    Park, J.H., Shin, J., Fung, P.: Reducing gender bias in abusive language detection. In: Pro- ceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. pp. 2799–2804. Association for Computational Linguistics, Brussels, Belgium (Oct-Nov 2018). https:/...

    2018 doi

  19. [26]

    In: Chinese Computational Linguistics and Natural Language Processing Based on Naturally Annotated Big Data, pp

    Qiu, Y ., Li, H., Li, S., Jiang, Y ., Hu, R., Yang, L.: Revisiting correlations between intrinsic and extrinsic evaluations of word embeddings. In: Chinese Computational Linguistics and Natural Language Processing Based on Naturally Annotated Big Data, pp. 209–221. Springer (2018)

    2018

  20. [27]

    arXiv preprint arXiv:1804.09301 (2018)

    Rudinger, R., Naradowsky, J., Leonard, B., Van Durme, B.: Gender bias in coreference resolution. arXiv preprint arXiv:1804.09301 (2018)

    2018 arXiv

  21. [28]

    In: Proceedings of the Third Workshop on Computational Typology and Multilingual NLP

    Sahai, S., Sharma, D.: Predicting and explaining french grammatical gender. In: Proceedings of the Third Workshop on Computational Typology and Multilingual NLP. pp. 90–96 (2021)

    2021

  22. [29]

    CoRR abs/1706.09799 (2017), http://arxiv.org/abs/1706.09799

    Sharma, S., El Asri, L., Schulz, H., Zumer, J.: Relevance of unsupervised metrics in task- oriented dialogue for evaluating natural language generation. CoRR abs/1706.09799 (2017), http://arxiv.org/abs/1706.09799

    2017 arXiv

  23. [30]

    arXiv preprint arXiv:2206.04615 (2022)

    Srivastava, A., Rastogi, A., Rao, A., Shoeb, A.A.M., Abid, A., Fisch, A., Brown, A.R., Santoro, A., Gupta, A., Garriga-Alonso, A., et al.: Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv preprint arXiv:2206.04615 (2022)

    2022 arXiv

  24. [31]

    arXiv preprint arXiv:1906.08976 (2019)

    Sun, T., Gaut, A., Tang, S., Huang, Y ., ElSherief, M., Zhao, J., Mirza, D., Belding, E., Chang, K.W., Wang, W.Y .: Mitigating gender bias in natural language processing: Literature review. arXiv preprint arXiv:1906.08976 (2019)

    2019 arXiv

  25. [32]

    In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics

    Sun, T., Gaut, A., Tang, S., Huang, Y ., ElSherief, M., Zhao, J., Mirza, D., Belding, E., Chang, K.W., Wang, W.Y .: Mitigating gender bias in natural language processing: Literature review. In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguis...

    2019 doi

  26. [33]

    arXiv preprint arXiv:2412.01253 (2024)

    Wake, A., Chen, B., Lv, C., Li, C., Huang, C., Cai, C., Zheng, C., Cooper, D., Zhou, F., Hu, F., et al.: Yi-lightning technical report. arXiv preprint arXiv:2412.01253 (2024)

    2024 arXiv

  27. [34]

    kelly is a warm person, joseph is a role model

    Wan, Y ., Pu, G., Sun, J., Garimella, A., Chang, K.W., Peng, N.: " kelly is a warm person, joseph is a role model": Gender biases in llm-generated reference letters. arXiv preprint arXiv:2310.09219 (2023) 12 Y . Li et al

    2023 arXiv

  28. [35]

    In: The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track (2024)

    Wang, Y ., Ma, X., Zhang, G., Ni, Y ., Chandra, A., Guo, S., Ren, W., Arulraj, A., He, X., Jiang, Z., et al.: Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. In: The Thirty-eight Conference on Neural Information Processing Systems Datasets ...

    2024

  29. [36]

    In: Proceedings of the First Workshop on NLP and Computational Social Science

    Waseem, Z.: Are you a racist or am I seeing things? annotator influence on hate speech detection on Twitter. In: Proceedings of the First Workshop on NLP and Computational Social Science. pp. 138–142. Association for Computational Linguistics, Austin, Texas (Nov 2016). https:/...

    2016 doi

  30. [37]

    Transactions of the Association for Computational Linguistics 6, 605–617 (2018)

    Webster, K., Recasens, M., Axelrod, V ., Baldridge, J.: Mind the gap: A balanced corpus of gendered ambiguous pronouns. Transactions of the Association for Computational Linguistics 6, 605–617 (2018)

    2018

  31. [38]

    Transactions of the Association for Computational Linguistics 6, 605–617 (2018)

    Webster, K., Recasens, M., Axelrod, V ., Baldridge, J.: Mind the GAP: A balanced corpus of gendered ambiguous pronouns. Transactions of the Association for Computational Linguistics 6, 605–617 (2018). https://doi.org/10.1162/tacl_a_00240, https://aclanthology.org/Q18-1042

    2018 doi

  32. [39]

    In: Chinese Corpus Linguistic Journal (Jan 2019)

    Weidong, Z., Rui, G., Baobao, C., Yirong, C., Chen, L.: Development of Peking University CCL Corpus. In: Chinese Corpus Linguistic Journal (Jan 2019)

    2019

  33. [40]

    PloS one 14(11), e0225385 (2019)

    Xu, H., Zhang, Z., Wu, L., Wang, C.J.: The cinderella complex: Word embeddings reveal gender stereotypes in movies and books. PloS one 14(11), e0225385 (2019)

    2019

  34. [41]

    Young, A., Chen, B., Li, C., Huang, C., Zhang, G., Zhang, G., Wang, G., Li, H., Zhu, J., Chen, J., et al.: Yi: Open foundation models by 01. ai. arXiv preprint arXiv:2403.04652 (2024)

    2024 arXiv

  35. [43]

    arXiv preprint arXiv:2405.19327 (2024)

    Zhang, G., Qu, S., Liu, J., Zhang, C., Lin, C., Yu, C.L., Pan, D., Cheng, E., Liu, J., Lin, Q., et al.: Map-neo: Highly capable and transparent bilingual large language model series. arXiv preprint arXiv:2405.19327 (2024)

    2024 arXiv

  36. [44]

    arXiv preprint arXiv:1905.07129 (2019)

    Zhang, Z., Han, X., Liu, Z., Jiang, X., Sun, M., Liu, Q.: Ernie: Enhanced language representa- tion with informative entities. arXiv preprint arXiv:1905.07129 (2019)

    2019 arXiv

  37. [45]

    In: Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing

    Zhao, J., Wang, T., Yatskar, M., Ordonez, V ., Chang, K.W.: Men also like shopping: Reducing gender bias amplification using corpus-level constraints. In: Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. pp. 2979–

    2017

  38. [46]

    In: Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, V olume 2 (Short Papers)

    Zhao, J., Wang, T., Yatskar, M., Ordonez, V ., Chang, K.W.: Gender bias in coreference resolution: Evaluation and debiasing methods. In: Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologi...

    2018 doi

  39. [47]

    arXiv preprint arXiv:1809.01496 (2018)

    Zhao, J., Zhou, Y ., Li, Z., Wang, W., Chang, K.W.: Learning gender-neutral word embeddings. arXiv preprint arXiv:1809.01496 (2018)

    2018 arXiv

  40. [48]

    In: Proceedings of the 20th Chinese National Conference on Computational Linguistics

    Zhao, J., Du, B., Zhu, S., Liu, P.: Construction of chinese sentence-level gender-unbiased data set and evaluation of gender bias in pre-training language. In: Proceedings of the 20th Chinese National Conference on Computational Linguistics. pp. 564–575 (2021)

    2021

  41. [49]

    arXiv preprint arXiv:1909.02224 (2019)

    Zhou, P., Shi, W., Zhao, J., Huang, K.H., Chen, M., Cotterell, R., Chang, K.W.: Examining gender bias in languages with grammatical gender. arXiv preprint arXiv:1909.02224 (2019)

    2019 arXiv

  42. [50]

    In: Proceedings of the 19th Chinese National Conference on Computational Linguistics

    Zhu, S., Liu, P.: Great males and stubborn females: A diachronic study of corpus-based gendered skewness in chinese adjectives. In: Proceedings of the 19th Chinese National Conference on Computational Linguistics. pp. 31–42 (2020)

    2020

  43. [51]

    In: Proceedings of the 19th Chinese National Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge 13 Conference on Computational Linguistics

    Zhu, S., Liu, P.: Great males and stubborn females: A diachronic study of corpus-based gendered skewness in Chinese adjectives. In: Proceedings of the 19th Chinese National Overview of the NLPCC 2025 Shared Task: Gender Bias Mitigation Challenge 13 Conference on Computational ...

    2025

  44. [2989]

    https: //doi.org/10.18653/v1/D17-1323, https://aclanthology.org/D17-1323

    Association for Computational Linguistics, Copenhagen, Denmark (Sep 2017). https: //doi.org/10.18653/v1/D17-1323, https://aclanthology.org/D17-1323

    2017 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.