Bias and Fairness in Large Language Models: A Survey
Open accessThe DOI matches its Crossref record; 308 of the 308 references it cites were found in a registry or an index; 151 of the 183 sources its claims cite were checked for support.
Abstract
Rapid advancements of large language models (LLMs) have enabled the processing, understanding, and generation of human-like text, with increasing integration into systems that touch our social sphere. Despite this success, these models can learn, perpetuate, and amplify harmful social biases. In this article, we present a comprehensive survey of bias evaluation and mitigation techniques for LLMs. We first consolidate, formalize, and expand notions of social bias and fairness in natural language processing, defining distinct facets of harm and introducing several desiderata to operationalize fairness for LLMs. We then unify the literature by proposing three intuitive taxonomies, two for bias evaluation, namely, metrics and datasets, and one for mitigation. Our first taxonomy of metrics for bias evaluation disambiguates the relationship between metrics and evaluation datasets, and organizes metrics by the different levels at which they operate in a model: embeddings, probabilities, and generated text. Our second taxonomy of datasets for bias evaluation categorizes datasets by their structure as counterfactual inputs or prompts, and identifies the targeted harms and social groups; we also release a consolidation of publicly available datasets for improved access. Our third taxonomy of techniques for bias mitigation classifies methods by their intervention during pre-processing, in-training, intra-processing, and post-processing, with granular subcategories that elucidate research trends. Finally, we identify open problems and challenges for future work. Synthesizing a wide range of recent research, we aim to provide a clear guide of the existing literature that empowers researchers and practitioners to better understand and prevent the propagation of bias in LLMs.
Overview
CoChat Summary
TL;DR
This survey consolidates, formalizes, and taxonomizes bias evaluation and mitigation for large language models, covering metrics, datasets, and mitigation techniques. It proposes three taxonomies and identifies open problems around power imbalances, fairness conceptualization, evaluation standards, mitigation scalability, and theoretical guarantees.
Key findings
- The survey proposes three taxonomies: bias evaluation metrics organized by what they use from a model (embeddings, probabilities, generated text); bias evaluation datasets organized by structure (counterfactual inputs vs. prompts); and bias mitigation techniques organized by intervention stage (pre-processing, in-training, intra-processing, post-processing).
- It formalizes social bias for LLMs as representational harms (e.g., derogatory language, disparate system performance, erasure, exclusionary norms, misrepresentation, stereotyping, toxicity) and allocational harms (direct and indirect discrimination), and introduces fairness desiderata including fairness through unawareness, invariance, equal social group associations, equal neutral associations, and replicated distributions.
- It reports limitations across metric classes: embedding-based and probability-based metrics may be only weakly or inconsistently correlated with downstream task bias, and generated-text metrics can be sensitive to decoding parameters; it also highlights reliability and validity issues in widely used datasets.
- It consolidates publicly available datasets, with reported sizes including CrowS-Pairs (1,508), StereoSet (16,995), BBQ (58,492), RealToxicityPrompts (100,000), BOLD (23,679), HolisticBias (460,000), and Bias NLI (5,712,066).
Methods
- Design: a comprehensive survey and taxonomy-building effort rather than an empirical study, synthesizing prior work on bias evaluation and mitigation for LLMs.
- Scope: focuses on bias issues in LLMs for English, with additional languages for machine translation and multilingual models, and restricts inclusion to works that propose novel closed-form metrics, datasets, or mitigation techniques and specify applicability to social bias or fairness.
- Analysis: organizes metrics, datasets, and mitigation techniques into taxonomies, formalizes several metrics and mitigation objectives with unified notation, discusses limitations of each class, and outlines open problems and recommendations.
Limitations
- The paper states that it aligns with a technical solutionist perspective rather than a critical theoretical one, and that its taxonomies are organized by technical implementation details instead of downstream usage contexts or harms.
- The paper states that it focuses narrowly on model training and evaluation, rather than a more holistic pipeline including problem formulation, data collection, and deployment/integration.
- The paper states that it is limited to English language papers; evident from the text, it also does not report a systematic search protocol, search dates, or screening counts.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
121 Valid · 12 Not valid · 18 Not checked · 106 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Despite the growing emphasis on addressing these issues, bias and fairness research in LLMs often fails to precisely describe the harms of model behaviors: who is harmed, why the behavior is harmful, and how the harm reflects and reinforces social principles or hierarchies (Blodgett et al. 2020).”
What the sources say
- Partly supportedcites Blodgett et al. 2020 sourceExperimental — accuracy not yet measured
Section 3.1 (Motivations; page not provided) finds that papers on bias in NLP often fail to explain which model behaviors are harmful, how, to whom, and why. Sections 4.1–4.2 call for attention to social hierarchies and explicit normative reasoning. The survey concerns NLP research broadly, however, and does not establish that this failure is common specifically in LLM research. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Similar to the shortcomings of embedding-based metrics, Delobelle et al. (2022) and Kaneko, Bollegala, and Okazaki (2022) point out that probability-based metrics may be only weakly correlated with biases that appear in downstream tasks”
- Also stated in: § 3.6 Recommendations · p. 27 · plain
“Probability-based metrics also show weak correlations with downstream biases.”
What the sources say
- Supportedcites Delobelle et al. 2022 source
Section 3 describes CrowS-Pairs as a pseudo-loglikelihood-based metric. Section 4.4 reports that CrowS-Pairs correlates less with two measures of bias in downstream tasks, supporting the hedged claim that probability-based metrics may correlate weakly with downstream bias. The finding does not establish that all probability-based metrics do so. (LLM, conf=0.90)
- Supportedcites Kaneko, Bollegala, and Okazaki 2022 source
The Introduction (p. 1) identifies intrinsic measures as using likelihood scores assigned by MLMs and reports only weak correlations with downstream, task-specific bias measures. Section 5.2 (p. 8) likewise finds weak and inconsistent correlations between intrinsic and extrinsic bias scores. (LLM, conf=0.93)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Also stated in: § 3.6 Recommendations · p. 27 · plain
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Many approaches, for instance, assume some implicitly desirable criterion ( e.g. , a model output should be independent of any social group in the input), but do not explicitly acknowledge or state the normative social values that justify their framework.”
What the sources say
- Partly supportedcites Blodgett et al. 2021 sourceExperimental — accuracy not yet measured
Section 2.3 describes assumptions about ideal model behavior, including equal probabilities for sentences about different target groups. Section 6 says the benchmarks make implicit decisions about stereotyping that are not explicitly discussed or justified. The paper examines four stereotyping benchmarks, however, rather than establishing this claim about many bias and fairness approaches generally or specifically about their normative social values. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Cabello, Jørgensen, and Søgaard (2023) illustrate that associations between the representations of protected attribute and other words can be independent of downstream performance disparities, if certain assumptions of social groups' language use are violated.”
What the sources say
- Supportedcites Cabello, Jørgensen, and Søgaard 2023 source
The Bias section defines association bias as differences in how demographic-group terms are encoded and empirical fairness as performance differences across social groups. Section 5 says the two can be independent and explains that linking them relies on the In-Group Affinity Assumption about how groups use those terms; it calls that assumption problematic. No page numbers are provided in our copy. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Each stakeholder brings perspectives that may specify different fairness constraints for the same application and setting.”
What the sources say
- Partly supportedcites Jernite et al. (2022) sourceExperimental — accuracy not yet measured
Section 2.2 (p. 4) says participants bring different perspectives and warns that relying only on their values could exclude views more relevant to other language users. The paper discusses nondiscrimination and local expressions of values, but does not say that stakeholders specify different fairness constraints for the same application and setting. (LLM, conf=0.85)
- Not matchedcites Bender (2019)
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
71 more claims · 34 with their cited sources checked
Create a free account to see all 257 claims and every source verdict.
- Probably not a substantial claim. 151
- It may state two separate claims as one. 70
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“While this process mitigates some risks inherent to AI-generated content, it does not eliminate potential biases associated with proprietary large language models, which lack transparency regarding training data and internal parameters [10].”
From AI-Generated Multiple Mini Interview (MMI) Stations for Medical School Admissions: Psychometric Evaluation (2026) · cites this paper as [10]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Notably, there is growing attention on assessing LLM bias [63-64] and robustness [323-324].”
From A Survey of Large Language Models (2026) · cites this paper as [63]
The cited source says what the claim attributes to it.
In the citing paper
“these models pose serious risks, as the vast, unfiltered internet data used to train them can reinforce and amplify harmful biases, disproportionately harming marginalized communities (73).”
From Emergent social conventions and collective bias in LLM populations (2025) · cites this paper as [73]
The cited source says what the claim attributes to it.
In the citing paper
“Despite the use of various mitigation techniques, such as debiasing algorithms, curating balanced datasets, or incorporating fairness-focused training objectives, eliminating bias entirely is a persistent challenge141–143.”
From Current applications and challenges in large language models for patient care: a systematic review (2025) · cites this paper as [142]
The cited source says what the claim attributes to it.
In the citing paper
“Another major concern with LLMs is their propensity to amplify historical biases.”
From Evaluation of Large Language Models for Peer Review in Transplantation Research: Algorithm Validation Study (2026) · cites this paper as [30]
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ 4.1 Counterfactual Inputs · p. 28
“| WinoBias+ | 1,367”
§ 4.1 Counterfactual Inputs · p. 29
“WinoBias+ (Vanmassenhove, Emmery, and Shterionov 2021) augments WinoBias with gender-neutral alternatives, similar to Winogender's neutral option, with 3,167 total instances.”
Citations
Cites
Items 201–312 of 312
MABEL: Attenuating Gender Bias using Textual Entailment Data
He, J., Xia, M., Fellbaum, C., & Chen, D. (2022) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2022.emnlp-main.657 · Cited by 60
Intra-Processing Methods for Debiasing Neural NetworksPreprint
Savani, Y., White, C., & Govindarajulu, N. S. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2006.08564 · Cited by 60 · cited 1×
“Following Savani et al. (2020)’s definition, we consider intra-processing methods to be those that take a pre-trained, perhaps fine-tuned, model as input, and modify the model’s behavior without further training or fine-tuning to generate debiased predictions at inference; as such, these techniques may also be conside…”
MethodDetoxifying Language Models Risks Marginalizing Minority Voices
Xu, A., Pathak, E., Wallace, E., Gururangan, S., Sap, M., & Klein, D. (2021) · Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2021.naacl-main.190 · Cited by 58
A Trip Towards Fairness: Bias and De-Biasing in Large Language Models
Ranaldi, L., Ruzzetti, E. S., Venditti, D., Onorati, D., & Zanzotto, F. M. (2024) · Joint Conference on Lexical and Computational Semantics · doi:10.18653/v1/2024.starsem-1.30 · Cited by 58
Demographic-Aware Language Model Fine-tuning as a Bias Mitigation Technique
Garimella, A., Mihalcea, R. F., & Amarnath, A. (2022) · Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the International Joint Conference on Natural Language Processing (AACL-IJCNLP) · doi:10.18653/v1/2022.aacl-short.38 · Cited by 58
UNQOVERing Stereotyping Biases via Underspecified Questions
Li, T., Khashabi, D., Khot, T., Sabharwal, A., & Srikumar, V. (2020) · Findings of the Association for Computational Linguistics: EMNLP 2020 · doi:10.18653/v1/2020.findings-emnlp.311 · Cited by 57
Delobelle, P., Tokpo, E., Calders, T., & Berendt, B. (2022) · Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2022.naacl-main.122 · Cited by 57 · cited 1×
“We also note that Delobelle et al. (2022) find it useful to differentiate between bias in the pre-trained model called intrinsic bias and bias that arises in the fine-tuning for a specific downstream task called extrinsic bias.”
BackgroundA Novel Estimator of Mutual Information for Learning to Disentangle Textual Representations
Colombo, P., Piantanida, P., & Clavel, C. (2021) · Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2021.acl-long.511 · Cited by 56
SaFeRDialogues: Taking Feedback Gracefully after Conversational Safety Failures
Ung, M., Xu, J., & Boureau, Y.-L. (2022) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2022.acl-long.447 · Cited by 56
Entropy-based attention regularization frees unintended bias mitigation from lists
Attanasio, G., Nozza, D., Hovy, D., & Baralis, E. (2022) · Findings of the Association for Computational Linguistics: ACL 2022 · doi:10.18653/v1/2022.findings-acl.88 · Cited by 56
Mei, K., Fereidooni, S., & Caliskan, A. (2023) · 2023 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3593013.3594109 · Cited by 56 · cited 2×
“, people with disabilities, mental illness, or low socioeconomic status) as negative (Mei et al., 2023).”
BackgroundClick: Controllable Text Generation with Sequence Likelihood Contrastive Learning
Zheng, C., Ke, P., Zhang, Z., & Huang, M. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.65 · Cited by 55
Li, Y., Du, M., Wang, X., & Wang, Y. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.797 · Cited by 54
Mitigating Gender Bias Amplification in Distribution by Posterior Regularization
Jia, S., Meng, T., Zhao, J., & Chang, K.-W. (2020) · Annual Meeting of the Association for Computational Linguistics · doi:10.48550/arxiv.2005.06251 · Cited by 53
How Gender Debiasing Affects Internal Model Representations, and Why It Matters
Orgad, H., Goldfarb-Tarrant, S., & Belinkov, Y. (2022) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2022.naacl-main.188 · Cited by 52
Social-Group-Agnostic Bias Mitigation via the Stereotype Content Model
Omrani, A., Ziabari, A. S., Yu, C., Golazizian, P., Kennedy, B., Atari, M., Ji, H., & Dehghani, M. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.227 · Cited by 52
Detecting Gender Stereotypes: Lexicon vs. Supervised Learning Methods
Cryan, J., Tang, S., Zhang, X., Metzger, M. J., Zheng, H.-T., & Zhao, B. Y. (2020) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376488 · Cited by 50
Theories of ”gender” in NLP bias research
Devinney, H., Björklund, J., & Björklund, H. (2022) · 2022 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3531146.3534627 · Cited by 49 · cited 2×
“For example, Devinney et al. (2022) argue that modeling gender as binary and immutable erases the identities of trans, nonbinary, and intersex people.”
BackgroundThe GEM Benchmark: Natural Language Generation, its Evaluation and Metrics
Gehrmann, S., Adewumi, T., Aggarwal, K., Ammanamanchi, P. S., Anuoluwapo, A., Bosselut, A., Chandu, K. R., Clinciu, M., Das, D., Dhole, K. D., et al. (2021) · arXiv · doi:10.48550/arxiv.2102.01672 · Cited by 49
TrustGPT: A Benchmark for Trustworthy and Responsible Large Language Models
Huang, Y., Zhang, Q., Y, P. S., & Sun, L. (2023) · arXiv · doi:10.48550/arxiv.2306.11507 · Cited by 47 · cited 4×
“Adapting INLP to a non-linear classifer, Iskander et al. (2023) proposes Iterative Gradient-Based Projection (IGBP), which leverages the gradients of a neural protected attribute classifier to project representations to the classifier’s class boundary, which should make the representations indistinguishable with respe…”
BackgroundInfluentialUnmasking the Mask – Evaluating Social Biases in Masked Language Models
Kaneko, M., & Bollegala, D. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i11.21453 · Cited by 46
Silva, A., Tambwekar, P., & Gombolay, M. (2021) · Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2021.naacl-main.189 · Cited by 46 · cited 1×
“During distilation, the student model may inherit or even amplify biases from the teacher (Ahn et al., 2022; Silva et al., 2021).”
BackgroundPruning for Protection: Increasing Jailbreak Resistance in Aligned LLMs Without Fine-Tuning
Hasan, A., Rugina, I., & Wang, A. (2024) · BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP (BlackboxNLP) · doi:10.18653/v1/2024.blackboxnlp-1.26 · Cited by 45
Challenges in Measuring Bias via Open-Ended Language Generation
Akyürek, A. F., Kocyigit, M. Y., Paik, S., & Wijaya, D. T. (2022) · GEBNLP · doi:10.48550/arxiv.2205.11601 · Cited by 44
Deep Learning on a Healthy Data Diet: Finding Important Examples for Fairness
Zayed, A., Parthasarathi, P., Mordido, G., Palangi, H., Shabanian, S., & Chandar, S. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i12.26706 · Cited by 43
BOLT: Fast Energy-based Controlled Text Generation with Tunable Biases
Liu, X., Khalifa, M., & Wang, L. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-short.18 · Cited by 43
Black, E., Naidu, R. C. A., Ghani, R., Rodolfa, K. T., Ho, D. E., & Heidari, H. (2023) · ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (EAAMO) · doi:10.1145/3617694.3623259 · Cited by 43
Detect and Perturb: Neutral Rewriting of Biased and Sensitive Text via Gradient-based Decoding
He, Z., Majumder, B. P., & McAuley, J. (2021) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2021.findings-emnlp.352 · Cited by 42
Steed, R., Panda, S., Kobren, A., & Wick, M. (2022) · Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2022.acl-long.247 · Cited by 42 · cited 1×
“Several works point out that biases in the embedding space have only weak or inconsistent relationships with biases in downstream tasks (Cabello et al., 2023; Cao et al., 2022; Goldfarb-Tarrant et al., 2021; Orgad & Belinkov, 2022; Orgad et al., 2022; Steed et al., 2022).”
BackgroundDiverse Adversaries for Mitigating Bias in Training
Han, X., Baldwin, T., & Cohn, T. (2021) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2021.eacl-main.239 · Cited by 41
Collecting a large-scale gender bias dataset for coreference resolution and machine translation
Levy, S., Lazar, K., & Stanovsky, G. (2021) · Findings of the Association for Computational Linguistics: EMNLP 2021 · doi:10.18653/v1/2021.findings-emnlp.211 · Cited by 41
On Transferability of Bias Mitigation Effects in Language Model Fine-Tuning
Jin, X., Barbieri, F., Kennedy, B., Mostafazadeh Davani, A., Neves, L., & Ren, X. (2021) · Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2021.naacl-main.296 · Cited by 40 · cited 1×
“Other works have since followed this framework (Han et al., 2021b; Jin et al., 2021), training an encoder and discriminator, where the discriminator predicts a protected attribute from a”
BackgroundBLIND: Bias Removal With No Demographics
Orgad, H., & Belinkov, Y. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.490 · Cited by 39
MoralDial: A Framework to Train and Evaluate Moral Dialogue Systems via Moral Discussions
Sun, H., Zhang, Z., Mi, F., Wang, Y., Liu, W., Cui, J., Wang, B., Liu, Q., & Huang, M. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.123 · Cited by 39
“Nice Try, Kiddo”: Investigating Ad Hominems in Dialogue Responses
Sheng, E., Chang, K.-W., Natarajan, P., & Peng, N. (2021) · Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2021.naacl-main.60 · Cited by 38
Choose Your Lenses: Flaws in Gender Bias Evaluation
Orgad, H., & Belinkov, Y. (2022) · Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.17 · Cited by 38 · cited 1×
“Peng et al. (2020) develop a reinforcement learning framework for fine-tuning to mitigate non-normative (i.”
BackgroundExploiting Biased Models to De-bias Text: A Gender-Fair Rewriting Model
Amrhein, C., Schottmann, F., Sennrich, R., & Läubli, S. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.246 · Cited by 38
Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey
Kumar, S., Balachandran, V., Njoo, L., Anastasopoulos, A., & Tsvetkov, Y. (2023) · Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.241 · Cited by 38
Parameter-efficient Modularised Bias Mitigation via AdapterFusion
Kumar, D., Lesota, O., Zerveas, G., Cohen, D., Eickhoff, C., Schedl, M., & Rekabsaz, N. (2023) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.201 · Cited by 37
Causally Testing Gender Bias in LLMs: A Case Study on Occupational Bias
Chen, Y., Raghuram, V. C., Mattern, J., Mihalcea, R. F., & Jin, Z. (2025) · Findings of the Association for Computational Linguistics: NAACL · doi:10.18653/v1/2025.findings-naacl.281 · Cited by 37
Iskander, S., Radinsky, K., & Belinkov, Y. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.369 · Cited by 36
Vanmassenhove, E., Emmery, C., & Shterionov, D. (2021) · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2021.emnlp-main.704 · Cited by 35 · cited 2×
“To provide gender-neutral alternatives to sentences with gendered pronouns, several works (Jain et al., 2021; Sun et al., 2021; Vanmassenhove et al., 2021) use a rules-based approach to generate parallel debiased sentences from biased sources, and then train a machine translation model to translate from biased sentenc…”
MethodGligorić, K., Cheng, M., Zheng, L., Durmus, E., & Jurafsky, D. (2024) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2024.naacl-long.331 · Cited by 33
Intrinsic bias metrics do not correlate with application bias
Goldfarb-Tarrant, S., Marchant, R., Muñoz Sánchez, R., Pandya, M., & Lopez, A. (2021) · Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) · doi:10.18653/v1/2021.acl-long.150 · Cited by 33
Joniak, P., & Aizawa, A. (2022) · Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.6 · Cited by 32
FairDistillation: Mitigating Stereotyping in Language Models
Delobelle, P., & Berendt, B. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-26390-3_37 · Cited by 32
Never Too Late to Learn: Regularizing Gender Bias in Coreference Resolution
Park, S., Choi, K., Yu, H., & Ko, Y. (2023) · ACM International Conference on Web Search and Data Mining (WSDM) · doi:10.1145/3539597.3570473 · Cited by 31 · cited 1×
“To mitigate gender stereotypes in occupation terms, Park et al. (2023) introduce a regularization term that orthogonalizes stereotypical word embeddings w and the gender direction g in the embedding space.”
BackgroundGrep-BiasIR: A dataset for investigating gender representation bias in information retrieval results
Krieg, K., Parada-Cabaleiro, E., Medicus, G., Lesota, O., Schedl, M., & Rekabsaz, N. (2023) · Proceedings of the 2023 Conference on Human Information Interaction and Retrieval · doi:10.1145/3576840.3578295 · Cited by 31 · cited 1×
“With a related task, Gender Representation-Bias for Information Retrieval (Grep-BiasIR) (Krieg et al., 2023) provides 118 gender-neutral search queries for document retrieval to assess gender representation bias.”
MethodModular and On-demand Bias Mitigation with Attribute-Removal Subnetworks
Hauzenberger, L., Masoudian, S., Kumar, D., Schedl, M., & Rekabsaz, N. (2023) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2023.findings-acl.386 · Cited by 31
Mitigating Gender Bias Amplification in Distribution by Posterior Regularization
Jia, S., Meng, T., Zhao, J., & Chang, K.-W. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.264 · Cited by 30
Joniak, P. K., & Aizawa, A. (2022) · GEBNLP · doi:10.48550/arxiv.2207.02463 · Cited by 30
Dynamically Disentangling Social Bias from Task-Oriented Representations with Adversarial Attack
Wang, L., Yan, Y., He, K., Wu, Y., & Xu, W. (2021) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2021.naacl-main.293 · Cited by 29 · cited 2×
“Wang et al. (2023) propose a fairness loss over the hidden states of the encoder to minimize the mutual information between the social group of a sentence (e.g., gender) and the sentence semantics (e.g., occupation). Similarly, Colombo et al. (2021) introduce a regularization term (Equation 37) to minimize mutual info…”
BackgroundLooking for a handsome carpenter! Debiasing GPT-3 job advertisements
Borchers, C., Gala, D., Gilburt, B., Oravkin, E., Bounsi, W., Asano, Y. M., & Kirk, H. (2022) · Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.22 · Cited by 29
Zhou, F., Mao, Y., Yu, L., Yang, Y., & Zhong, T. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.acl-long.232 · Cited by 29
On the challenges of using black-box APIs for toxicity evaluation in research
Pozzobon, L., Ermis, B., Lewis, P., & Hooker, S. (2023) · Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2023.emnlp-main.472 · Cited by 29
Li, Y., Zhang, L., & Zhang, Y. (2023) · arXiv · doi:10.48550/arxiv.2305.18569 · Cited by 28
They, Them, Theirs: Rewriting with Gender-Neutral English
Sun, T., Webster, K., Shah, A., Wang, W. Y., & Johnson, M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2102.06788 · Cited by 26 · cited 1×
“To provide gender-neutral alternatives to sentences with gendered pronouns, several works (Jain et al., 2021; Sun et al., 2021; Vanmassenhove et al., 2021) use a rules-based approach to generate parallel debiased sentences from biased sources, and then train a machine translation model to translate from biased sentenc…”
MethodSocially Aware Bias Measurements for Hindi Language Representations
Malik, V., Dev, S., Nishi, A., Peng, N., & Chang, K.-W. (2022) · Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT) · doi:10.18653/v1/2022.naacl-main.76 · Cited by 25
Theory-Grounded Measurement of U.S. Social Stereotypes in English Language Models
Cao, Y., Sotnikova, A., Daumé III, H., Rudinger, R., & Zou, L. (2022) · Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies · doi:10.18653/v1/2022.naacl-main.92 · Cited by 24
Improving Gender-Related Fairness in Sentence Encoders: A Semantics-Based Approach
Dolci, T., Azzalini, F., & Tanelli, M. (2023) · Data Science and Engineering · doi:10.1007/s41019-023-00211-0 · Cited by 23
Denton, E., Díaz, M., Kivlichan, I., Prabhakaran, V., & Rosen, R. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2112.04554 · Cited by 22
Learning to Generate Equitable Text in Dialogue from Biased Training Data
Sicilia, A., & Alikhani, M. (2023) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2023.acl-long.163 · Cited by 22
Controlling Bias Exposure for Fair Interpretable Predictions
He, Z., Wang, Y., McAuley, J., & Majumder, B. P. (2022) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2022.findings-emnlp.431 · Cited by 22
Decoupling Adversarial Training for Fair NLP
Han, X., Baldwin, T., & Cohn, T. (2021) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2021.findings-acl.41 · Cited by 21 · cited 9×
“For bias mitigation in a setting with only limited labeling of protected attributes, Han et al. (2021a) propose a modified optimization objective that separates discriminator training from the main model training, so that the discriminator can be selectively applied to only the instances with a social grouplabel. To m…”
BackgroundMethodInfluentialCompensatory Debiasing For Gender Imbalances In Language Models
Woo, T.-J., Nam, W.-J., Ju, Y.-J., & Lee, S. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10095658 · Cited by 21 · cited 3×
“L = ∑ i,j∈{1,··· ,d},i<j JS (P ai∥P aj ) + λKL (Q∥P ) (31) In their method GuiDebias, Woo et al. (2023) consider gender stereotype sentences, with a regularization term (Equation 32) to enforce independence between gender groups and the representations of stereotypical masculine Sm and feminine Sf sentences, given by…”
BackgroundMethodReducing Non-Normative Text Generation from Language Models
Peng, X., Li, S., Frazier, S., & Riedl, M. (2020) · Proceedings of the 13th International Conference on Natural Language Generation · doi:10.18653/v1/2020.inlg-1.43 · Cited by 20
When the Majority is Wrong: Modeling Annotator Disagreement for Subjective Tasks
Fleisig, E., Abebe, R., & Klein, D. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.415 · Cited by 20
Parameter-Efficient Transfer Learning with Diff Pruning
Guo, D., Rush, A. M., & Kim, Y. (2020) · arXiv · doi:10.48550/arxiv.2012.07463 · Cited by 20 · cited 3×
“R(t) = λ ∣∣∣∣∣log ∑|Ai| k=1 P (Ai,k)∑|Aj | k=1 P (Aj,k) ∣∣∣∣∣ (41) In Auto-Debias, Guo et al. (2022) extend these ideas to non-binary social groups, encouraging the generated output to be independent of social group.”
BackgroundMethodInfluentialImplicit attitudes and the perception of sociolinguistic variation
Loudermilk, B. C. (2015) · IMPACT: Studies in Language, Culture and Society · doi:10.1075/impact.39.06lou · Cited by 19 · cited 4×
“Counterfactual data augmentation (CDA) is one of the primary of these augmentation techniques (Lu et al., 2020; Qian et al., 2022; Webster et al., 2020; Zmigrod et al., 2019), replacing protected attribute words, such as gendered pronouns, to achieve a balanced dataset. In one of the first formalizations of this appro…”
BackgroundMethodInfluentialChallenges in measuring bias via open-ended language generation
Akyürek, A. F., Kocyigit, M. Y., Paik, S., & Wijaya, D. T. (2022) · Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.9 · Cited by 19
Don’t Just Clean It, Proxy Clean It: Mitigating Bias by Proxy in Pre-Trained Models
Panda, S., Kobren, A., Wick, M., & Shen, Q. (2022) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/v1/2022.findings-emnlp.372 · Cited by 19 · cited 1×
“Panda et al. (2022) present D-Bias to identify proxy words via co-occurence frequencies, and mask out identity words and their proxies prior to fine-tuning.”
BackgroundAhn, J., Lee, H., Kim, J., & Oh, A. (2022) · Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.27 · Cited by 18 · cited 5×
“During distilation, the student model may inherit or even amplify biases from the teacher (Ahn et al., 2022; Silva et al., 2021).”
BackgroundMethodInfluentialUnified Detoxifying and Debiasing in Language Generation via Inference-time Adaptive Optimization
Yang, Z., Yi, X., Li, P., Liu, Y., & Xie, X. (2022) · arXiv · doi:10.48550/arxiv.2210.04492 · Cited by 17 · cited 1×
“We provide a non-exhaustive list of examples of settings where bias may manifest in unique forms, depending on the task: • Text Generation: In generated text, bias may appear locally or globally (Liang et al., 2021; Sheng et al., 2019; Yang et al., 2022).”
BackgroundMixup-based Unified Framework to Overcome Gender Bias Resurgence
Liu, Y., Mao, Y., Wu, J., & Zhou, F. (2023) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3539618.3591938 · Cited by 17 · cited 1×
“Based on Zhang et al. (2018b)’s mixup technique, interpolation techniques interpolate counterfactually-augmented training examples with the original versions and their labels to extend the distribution of the training data.”
MethodGender-tuning: Empowering Fine-tuning for Debiasing Pre-trained Language Models
Ghanbarzadeh, S., Huang, Y., Palangi, H., Cruz Moreno, R., & Khanpour, H. (2023) · Findings of the Association for Computational Linguistics: ACL 2023 · doi:10.18653/v1/2023.findings-acl.336 · Cited by 17
Kaneko, M., Bollegala, D., & Okazaki, N. (2022) · arXiv · doi:10.48550/arxiv.2210.02938 · Cited by 17
First the Worst: Finding Better Gender Translations During Beam Search
Saunders, D., Sallis, R., & Byrne, B. (2022) · Findings of the Association for Computational Linguistics: ACL 2022 · doi:10.18653/v1/2022.findings-acl.301 · Cited by 16
HeteroCorpus: A Corpus for Heteronormative Language Detection
Vásquez, J., Bel-Enguix, G., Andersen, S., & Ojeda-Trueba, S.-L. (2022) · Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.23 · Cited by 16 · cited 2×
“(2020) and Xu et al. (2020) propose word- or n-gram blocking during decoding, prohibiting the use of tokens from an unsafe word list.”
BackgroundShould We Attend More or Less? Modulating Attention for FairnessPreprint
Zayed, A., Mordido, G., Shabanian, S., & Chandar, S. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2305.13088 · Cited by 16
A comparative study on the impact of model compression techniques on fairness in language models
Ramesh, K., Chavan, A., Pandit, S., & Sitaram, S. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.acl-long.878 · Cited by 15
Toward Fairness in Text Generation via Mutual Information Minimization based on Importance SamplingPreprint
Wang, R., Cheng, P., & Henao, R. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2302.13136 · Cited by 15
On measuring and mitigating biased inferences of word embeddings
Dev, S., Li, T., Phillips, J. M., & Srikumar, V. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i05.6267 · Cited by 14
Debiasing pretrained text encoders by paying attention to paying attention
Gaci, Y., Benatallah, B., Casati, F., & Benabdeslem, K. (2022) · Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2022.emnlp-main.651 · Cited by 14
Learn What NOT to Learn: Towards Generative Safety in ChatbotsPreprint
Khalatbari, L., Bang, Y., Su, D., Chung, W., Ghadimi, S., Sameti, H., & Fung, P. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2304.11220 · Cited by 14
On the independence of association bias and empirical fairness in language models
Cabello, L., Jørgensen, A. K., & Søgaard, A. (2023) · 2023 ACM Conference on Fairness, Accountability, and Transparency · doi:10.1145/3593013.3594004 · Cited by 14 · cited 6×
“Several works point out that biases in the embedding space have only weak or inconsistent relationships with biases in downstream tasks (Cabello et al., 2023; Cao et al., 2022; Goldfarb-Tarrant et al., 2021; Orgad & Belinkov, 2022; Orgad et al., 2022; Steed et al., 2022).”
BackgroundInfluentialFairPrism: Evaluating Fairness-Related Harms in Text Generation
Fleisig, E., Amstutz, A., Atalla, C., Blodgett, S. L., Daumé III, H., Olteanu, A., Sheng, E., Vann, D., & Wallach, H. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.acl-long.343 · Cited by 14 · cited 2×
“, 2022), for instance, contains examples tweets labeled as non-heteronormative, heteronormative to assess negative impacts on the LGBTQ+ community, and FairPrism (Fleisig et al., 2023) provides examples of stereotyping and derogatory biases with respect to gender and sexuality.”
BackgroundTowards Equal Opportunity Fairness through Adversarial LearningPreprint
Han, X., Baldwin, T., & Cohn, T. (2022) · arXiv.org · doi:10.48550/arxiv.2203.06317 · Cited by 14
Moral responsibility for computing artifacts
Grodzinsky, F. S., Miller, K., & Wolf, M. J. (2012) · ACM SIGCAS Computers and Society · doi:10.1145/2422509.2422511 · Cited by 13
InterFair: Debiasing with Natural Language Feedback for Fair Interpretable Predictions
Majumder, B. P., He, Z., & McAuley, J. (2023) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2023.emnlp-main.589 · Cited by 13
Using In-Context Learning to Improve Dialogue Safety
Meade, N., Gella, S., Hazarika, D., Gupta, P., Jin, D., Reddy, S., Liu, Y., & Hakkani-Tur, D. (2023) · Findings of the Association for Computational Linguistics: EMNLP 2023 · doi:10.18653/v1/2023.findings-emnlp.796 · Cited by 13
Limisiewicz, T., & Mareček, D. (2022) · Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.3 · Cited by 11
The tail wagging the dog: Dataset construction biases of social bias benchmarks
Selvam, N., Dev, S., Khashabi, D., Khot, T., & Chang, K.-W. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) · doi:10.18653/v1/2023.acl-short.118 · Cited by 11
Generating Gender Augmented Data for NLP
Jain, N., Popović, M., Groves, D., & Vanmassenhove, E. (2021) · Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2021.gebnlp-1.11 · Cited by 10 · cited 1×
“To provide gender-neutral alternatives to sentences with gendered pronouns, several works (Jain et al., 2021; Sun et al., 2021; Vanmassenhove et al., 2021) use a rules-based approach to generate parallel debiased sentences from biased sources, and then train a machine translation model to translate from biased sentenc…”
MethodSattigeri, P., Ghosh, S. K., Padhi, I., Dognin, P., & Varshney, K. R. (2022) · neural information processing systems · doi:10.52202/068431-2601 · Cited by 10
Causally Testing Gender Bias in LLMs: A Case Study on Occupational BiasPreprint
Chen, Y., Raghuram, V. C., Mattern, J., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2212.10678 · Cited by 9
Disrupting the Gospel of Tech Solutionism to Build Tech Justice
Byrum, G., & Benjamin, R. (2022) · Stanford Social Innovation Review · doi:10.48558/9sev-4d26 · Cited by 9
Fair Enough: Standardizing Evaluation and Model Selection for Fairness Research in NLP
Han, X., Baldwin, T., & Cohn, T. (2023) · Conference of the European Chapter of the Association for Computational Linguistics · doi:10.18653/v1/2023.eacl-main.23 · Cited by 8
Does Representational Fairness Imply Empirical Fairness?
Shen, A., Han, X., Cohn, T., Baldwin, T., & Frermann, L. (2022) · AACL/IJCNLP · doi:10.18653/v1/2022.findings-aacl.8 · Cited by 7
A Taxonomy of Bias-Causing Ambiguities in Machine Translation
Měchura, M. (2022) · Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.18 · Cited by 5 · cited 2×
“• Machine Translation: Machine translators may default to masculine words in the case of ambiguity, a form of an exclusionary norm, such as translating from English "I am happy" to the masculine French form "je suis heureux", as opposed to the feminine form "je suis heureuse" (Měchura, 2022).”
BackgroundWhat changed? Investigating debiasing methods using causal mediation analysis
Jeoung, S., & Diesner, J. (2022) · Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.26 · Cited by 5
Sattigeri, P., Ghosh, S. K., Padhi, I., et al. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2212.06803 · Cited by 5
Incorporating subjectivity into gendered ambiguous pronoun (GAP) resolution using style transfer
Pant, K., & Dadu, T. (2022) · Proceedings of the 4th Workshop on Gender Bias in Natural Language Processing (GeBNLP) · doi:10.18653/v1/2022.gebnlp-1.28 · Cited by 3
Sociodemographic Bias in Language Models: A Survey and Forward PathPreprint
Gupta, V. K., Venkit, P. N., Wilson, S., et al. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2306.08158 · Cited by 3 · cited 1×
“As we discuss in Section 4, several widely-used evaluation datasets suffer from reliability and validity issues, including ambiguities about whether instances accurately reflect real-world stereotypes, inconsistent treatment of social groups, assumptions of near-perfect understanding of language, and lack of syntactic…”
BackgroundPruning for Protection: Increasing Jailbreak Resistance in Aligned LLMs Without Fine-TuningPreprint
Hasan, A., Rugina, I., & Wang, A. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2401.10862 · Cited by 3
Pay Attention to Your Tone: Introducing a New Dataset for Polite Language RewritePreprint
Wang, X., Ge, T., Mao, A., Li, Y., Wei, F., & Chen, S. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2212.10190 · Cited by 2
Understanding Stereotypes in Language Models: Towards Robust Measurement and Zero-Shot DebiasingPreprint
Mattern, J., Jin, Z., Sachan, M., Mihalcea, R., & Schölkopf, B. (2022) · Repository for Publications and Research Data (ETH Zurich) · doi:10.3929/ethz-b-000653515 · Cited by 2
The other side of compression: Measuring bias in pruned transformers
Proskurina, I., Metzler, G., & Velcin, J. (2023) · Lecture Notes in Computer Science · doi:10.1007/978-3-031-30047-9_29 · Cited by 1
Gligorić, K., Cheng, M., Zheng, L., et al. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2404.01651
What Changed? Investigating Debiasing Methods using Causal Mediation AnalysisPreprint
Jeoung, S., & Diesner, J. (2022) · arXiv (Cornell University) · doi:10.48550/arxiv.2206.00701
A Trip Towards Fairness: Bias and De-Biasing in Large Language Models
Ranaldi, L., Ruzzetti, E. S., Venditti, D., et al. (2023) · arXiv · doi:10.48550/arxiv.2305.13862
Untitled work
Srivastava (2014) · doi:10.5555/2627435.2670313
OpenAI, Achiam, J., Adler, S., Agarwal, S., Ahmad, L., Akkaya, I., Aleman, F. L., Almeida, D., Altenschmidt, J., Altman, S., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.08774
Research works printed but not linked: 5
- Barocas, 2019, Fairness and Machine Learning: Limitations and Opportunities
- Blodgett, Su Lin . 2021. Sociolinguistically Driven Approaches for Just Natural Language Processing. Ph.D. thesis. University of Massachusetts Amherst.
- Green, 2019, ”Good” isn’t good enough, 1
- Radford, 2019, Language models are unsupervised multitask learners, OpenAI Blog, 1, 9
- Wang, 2019, BERT has a mouth, and it must speak: BERT as a Markov random field language model, 30, 10.18653/v1/w19-2304 · doi:10.18653/v1/w19-2304
Cited by
A Survey of Large Language Models
Zhao, W. X., Zhou, K., Li, J., Tang, T., Dong, Z., Hou, Y., Zhang, B., Min, Y., Zhang, J., Liu, P., et al. (2026) · Frontiers of Computer Science · doi:10.1007/s11704-026-60308-3 · Cited by 2,093
Gu, J., Jiang, X., Shi, Z., Tan, H., Zhai, X., Xu, C., Li, W., Shen, Y., Ma, S., Liu, H., et al. (2024) · arXiv · doi:10.48550/arxiv.2411.15594 · Cited by 1,781
“122 In the context of LLM-as-a-judge, it indicates that evaluators might score expressions from unfamiliar cultures poorly.”
A Survey on the Memory Mechanism of Large Language Model based Agents
Zhang, Z., Bo, X., Ma, C., Li, R., Chen, X., Dai, Q., Zhu, J., Dong, Z., & Wen, J.-R. (2024) · arXiv · doi:10.48550/arxiv.2404.13501 · Cited by 149
“Gödel Agent [206] utilizes dynamic code modification techniques to alter its own logic and behavior based on the feedback signals from the environment and stores the updated agent logic in the agent memory for recursive improvement.”
Security and Privacy Challenges of Large Language Models: A Survey
Das, B. C., Amini, M. H., & Wu, Y. (2025) · ACM Computing Surveys · doi:10.1145/3712001 · Cited by 445
A survey on LLM-based multi-agent systems: Workflow, infrastructure, and challenges
Li, X., Wang, S., Zeng, S., Wu, Y., & Yang, Y. (2024) · Vicinagearth · doi:10.1007/s44336-024-00009-2 · Cited by 404
“(3 ) Bias : Bias in LLMs manifests as the propagation and amplification of discriminatory tendencies present in training data, such as racial and gender biases, leading to unfair or harmful outputs [327–330].”
Metadata
- Authors
- Gallegos, I. O., Rossi, R. A., Barrow, J., Tanjim, M. M., Kim, S., Dernoncourt, F., Yu, T., Zhang, R., & Ahmed, N. K.
- Affiliations
- Stanford University (US)
- Adobe Research. ryrossi@adobe.com
- Pattern Data. joe.barrow@patterndataworks.com
- Adobe Research. tanjim@adobe.com
- Adobe Research. sukim@adobe.com
- Adobe Research. dernonco@adobe.com
- Adobe Research. tyu@adobe.com
- Adobe Research. ruizhang@adobe.com
- Intel (United States) (US)
- ORCID
- Venue
- Computational Linguistics
- Publisher
- MIT Press
- Published
- 2024
- Volume
- 50
- Issue
- 3
- Pages
- 1097-1179
- ISSN
- 0891-2017 (ISSN-L)
- 1530-9312
- OpenAlex
- W4399528455
- Semantic Scholar
- bcfa73aedf1b2d1ee4f168e21298a37ac55a37f7
- Cited by
- 1,245
- References
- 280
- Citations by year
- 2026: 596 · 2025: 451 · 2024: 188
- FWCI
- 132.76
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- Locations
- Computational Linguistics (publishedVersion, cc-by-nc-nd)
- arXiv (Cornell University) (submittedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Artificial Intelligence › Hate Speech and Cyberbullying Detection
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Keywords
- large language models, social bias, fairness, bias assessment, bias mitigation, bias metrics, data preprocessing, post-processing
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-11-02, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-02 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.