mlsanityData to 7 Oct 2026

Self-Debiasing Large Language Models: Zero-Shot Recognition and Reduction of Stereotypes

arXiv 2402.01981 ↗First posted 3 Feb 2024cs.CL cs.AI cs.CYcited by 58

Not on the map. Why?

Abstract (arXiv)

Large language models (LLMs) have shown remarkable advances in language generation and understanding but are also prone to exhibiting harmful social biases. While recognition of these behaviors has generated an abundance of bias mitigation techniques, most require modifications to the training data, model parameters, or decoding strategy, which may be infeasible without access to a trainable model. In this work, we leverage the zero-shot capabilities of LLMs to reduce stereotyping in a technique we introduce as zero-shot self-debiasing. With two approaches, self-debiasing via explanation and self-debiasing via reprompting, we show that self-debiasing can significantly reduce the degree of stereotyping across nine different social groups while relying only on the LLM itself and a simple prompt, with explanations correctly identifying invalid assumptions and reprompting delivering the greatest reductions in bias. We hope this work opens inquiry into other zero-shot techniques for bias mitigation.

Key facts. Cited by 58 papers in the map (data through 7 Oct 2026). Builds on 42 of its 74 references that are in the map. First posted on arXiv 3 Feb 2024.

Builds on

42 of 74 references.

References · 42

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2022Large Language Models are Zero-Shot Reasonerscited by 4,323
  3. 2021Calibrate Before Use: Improving Few-Shot Performance of Language Modelscited by 1,371
  4. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  5. 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 951
  6. 2019Plug and Play Language Models: A Simple Approach to Controlled Text Generationcited by 793
  7. 2021BBQ: A Hand-Built Bias Benchmark for Question Answeringcited by 683
  8. 2022Large Language Models Can Self-Improvecited by 656
  9. 2021Prompt Programming for Large Language Models: Beyond the Few-Shot Paradigmcited by 599
  10. 2021DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Expertscited by 434
  11. 2020GeDi: Generative Discriminator Guided Sequence Generationcited by 377
  12. 2021Persistent Anti-Muslim Bias in Large Language Modelscited by 376
30 more
  1. 2021Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLPcited by 336
  2. 2022Quark: Controllable Text Generation with Reinforced Unlearningcited by 236
  3. 2023LLM Self Defense: By Self Examination, LLMs Know They Are Being Trickedcited by 226
  4. 2020Measuring and Reducing Gendered Correlations in Pre-trained Modelscited by 222
  5. 2018Gender Bias in Neural Natural Language Processingcited by 210
  6. 2021Societal Biases in Language Generation: Progress and Challengescited by 201
  7. 2020Social Biases in NLP Models as Barriers for Persons with Disabilitiescited by 182
  8. 2019Queens are Powerful too: Mitigating Gender Bias in Dialogue Generationcited by 152
  9. 2020Towards Controllable Biases in Language Generationcited by 112
  10. 2021Debiasing Pre-trained Contextualised Embeddingscited by 107
  11. 2019Does Gender Matter? Towards Fairness in Dialogue Systemscited by 89
  12. 2021FairFil: Contrastive Neural Debiasing Method for Pretrained Text Encoderscited by 75
  13. 2019Reducing Gender Bias in Word-Level Language Models with a Gender-Equalizing Loss Functioncited by 57
  14. 2022Entropy-based Attention Regularization Frees Unintended Bias Mitigation from Listscited by 54
  15. 2023Bias Against 93 Stigmatized Groups in Masked Language Models and Downstream Sentiment Classification Taskscited by 51
  16. 2021NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternativescited by 46
  17. 2021They, Them, Theirs: Rewriting with Gender-Neutral Englishcited by 44
  18. 2023Using In-Context Learning to Improve Dialogue Safetycited by 41
  19. 2023Click: Controllable Text Generation with Sequence Likelihood Contrastive Learningcited by 41
  20. 2021Improving Gender Fairness of Pre-Trained Language Models without Catastrophic Forgettingcited by 34
  21. 2022Learning Fair Representation via Distributional Contrastive Disentanglementcited by 31
  22. 2022MABEL: Attenuating Gender Bias using Textual Entailment Datacited by 31
  23. 2021First the worst: Finding better gender translations during beam searchcited by 28
  24. 2022Causally Testing Gender Bias in LLMs: A Case Study on Occupational Biascited by 28
  25. 2020Mitigating Gender Bias Amplification in Distribution by Posterior Regularizationcited by 27
  26. 2022Text Style Transfer for Bias Mitigation using Masked Language Modelingcited by 27
  27. 2022Deep Learning on a Healthy Data Diet: Finding Important Examples for Fairnesscited by 21
  28. 2022Controlling Bias Exposure for Fair Interpretable Predictionscited by 18
  29. 2021Generating Gender Augmented Data for NLPcited by 11
  30. 2022InterFair: Debiasing with Natural Language Feedback for Fair Interpretable Predictionscited by 7

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 15; 2025: 28; 2026: 15.

58 papers.

Sort:

By area:

  1. 2024Social Bias Evaluation for Large Language Models Requires Prompt Variationscited by 39
  2. 2024Digital Forgetting in Large Language Models: A Survey of Unlearning Methodscited by 38
  3. 2024Prompting Techniques for Reducing Social Bias in LLMs through System 1 and System 2 Cognitive Processescited by 38
  4. 2025FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation Steeringcited by 37
  5. 2024MBBQ: A Dataset for Cross-Lingual Comparison of Stereotypes in Generative LLMscited by 32
  6. 2025Fairness through Difference Awareness: Measuring Desired Group Discrimination in LLMscited by 20
  7. 2024BiasDPO: Mitigating Bias in Language Models through Direct Preference Optimizationcited by 18
  8. 2024Walking in Others' Shoes: How Perspective-Taking Guides Large Language Models in Reducing Toxicity and Biascited by 17
  9. 2025A Survey of Foundation Model-Powered Recommender Systems: From Feature-Based, Generative to Agentic Paradigmscited by 17
  10. 2025Dissecting Bias in LLMs: A Mechanistic Interpretability Perspectivecited by 14
  11. 2024Social Bias in Large Language Models For Bangla: An Empirical Study on Gender and Religious Biascited by 13
  12. 2024Evaluating and Mitigating Social Bias for Large Language Models in Open-ended Settingscited by 8
38 more
  1. 2024Mitigating Social Bias in Large Language Models: A Multi-Objective Approach within a Multi-Agent Frameworkcited by 8
  2. 2025Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidancecited by 7
  3. 2025What's Not Said Still Hurts: A Description-Based Evaluation Framework for Measuring Social Bias in LLMscited by 6
  4. 2025More of the Same: Persistent Representational Harms Under Increased Representationcited by 6
  5. 2025BiasFilter: An Inference-Time Debiasing Framework for Large Language Modelscited by 6
  6. 2026CatRAG: Functor-Guided Structural Debiasing with Retrieval Augmentation for Fair LLMscited by 6
  7. 2024Unboxing Occupational Bias: Grounded Debiasing of LLMs with U.S. Labor Datacited by 5
  8. 2025C2PO: Diagnosing and Disentangling Bias Shortcuts in LLMscited by 5
  9. 2026Bi-directional Bias Attribution: Debiasing Large Language Models without Modifying Promptscited by 5
  10. 2024An Empirical Study of Gendered Stereotypes in Emotional Attributes for Bangla in Multilingual Large Language Modelscited by 4
  11. 2025Rethinking Prompt-based Debiasing in Large Language Modelscited by 4
  12. 2025BiasGuard: A Reasoning-enhanced Bias Detection Tool For Large Language Modelscited by 4
  13. 2024Different Bias Under Different Criteria: Assessing Bias in LLMs with a Fact-Based Approachcited by 3
  14. 2025Discourse Heuristics For Paradoxically Moral Self-Correctioncited by 3
  15. 2025Bias after Prompting: Persistent Discrimination in Large Language Modelscited by 3
  16. 2024REFINE-LM: Mitigating Language Model Stereotypes via Reinforcement Learningcited by 2
  17. 2025Purdah and Patriarchy: Evaluating and Mitigating South Asian Biases in Open-Ended Multilingual LLM Generationscited by 2
  18. 2025Smoothie-Qwen: Post-Hoc Smoothing to Reduce Language Bias in Multilingual LLMscited by 2
  19. 2025AMBEDKAR-A Multi-level Bias Elimination through a Decoding Approach with Knowledge Augmentation for Robust Constitutional Alignment of Language Modelscited by 2
  20. 2025Bias Mitigation or Cultural Commonsense? Evaluating LLMs with a Japanese Datasetcited by 2
  21. 2025BiasFreeBench: a Benchmark for Mitigating Bias in Large Language Model Responsescited by 2
  22. 2025Personalized Decision Modeling: Utility Optimization or Textualized-Symbolic Reasoningcited by 2
  23. 2026Self-Blinding and Counterfactual Self-Simulation Mitigate Biases and Sycophancy in Large Language Modelscited by 2
  24. 2026Implicit Bias in LLMs for Transgender Populationscited by 2
  25. 2026Self-Debias: Self-correcting for Debiasing Large Language Modelscited by 2
  26. 2024Agentic AI: The Era of Semantic Decodingcited by 1
  27. 2025Beyond Bias Scores: Unmasking Vacuous Neutrality in Small Language Modelscited by 1
  28. 2025BiasGym: A Simple and Generalizable Framework for Analyzing and Removing Biases through Injectioncited by 1
  29. 2025Mitigation of Gender and Ethnicity Bias in AI-Generated Stories through Model Explanationscited by 1
  30. 2025Let's Unlearn Stereotypes Before Decision-Making: Assessing the Impact of Intrinsic Bias Mitigation on Downstream Fairness in LLMscited by 1
  31. 2025Mitigating Judgment Preference Bias in Large Language Models through Group-Based Pollingcited by 1
  32. 2025Can SAEs reveal and mitigate racial biases of LLMs in healthcare?cited by 1
  33. 2025Cross-Language Bias Examination in Large Language Modelscited by 1
  34. 2026Identity-Robust Language Model Generation via Content Integrity Preservationcited by 1
  35. 2026Routing Sensitivity Without Controllability: A Diagnostic Study of Fairness in MoE Language Modelscited by 1
  36. 2024Debias your Large Multi-Modal Model at Test-Time via Non-Contrastive Visual Attribute Steeringcited by 0
  37. 2025DeCAP: Context-Adaptive Prompt Generation for Debiasing Zero-shot Question Answering in Large Language Modelscited by 0
  38. 2025Debiasing LLMs by Masking Unfairness-Driving Attention Headscited by 0

50 most cited of 58 All 58 (JSON)