mlsanityData to 7 Oct 2026

UnQovering Stereotyping Biases via Underspecified Questions

arXiv 2010.02428 ↗First posted 6 Oct 2020cs.CLcited by 124

Not on the map. Why?

Abstract (arXiv)

While language embeddings have been shown to have stereotyping biases, how these biases affect downstream question answering (QA) models remains unexplored. We present UNQOVER, a general framework to probe and quantify biases through underspecified questions. We show that a naive use of model scores can lead to incorrect bias estimates due to two forms of reasoning errors: positional dependence and question independence. We design a formalism that isolates the aforementioned errors. As case studies, we use this metric to analyze four important classes of stereotypes: gender, nationality, ethnicity, and religion. We probe five transformer-based QA models trained on two QA datasets, along with their underlying language models. Our broad study reveals that (1) all these models, with and without fine-tuning, have notable stereotyping biases in these classes; (2) larger models often have higher bias; and (3) the effect of fine-tuning on bias varies strongly with the dataset and the model size.

Key facts. Cited by 124 papers in the map (data through 7 Oct 2026). First posted on arXiv 6 Oct 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 12; 2022: 13; 2023: 31; 2024: 34; 2025: 26; 2026: 8.

124 papers.

Sort:

By area:

  1. 2021GLaM: Efficient Scaling of Language Models with Mixture-of-Expertscited by 890
  2. 2021BBQ: A Hand-Built Bias Benchmark for Question Answeringcited by 683
  3. 2023Bias and Fairness in Large Language Models: A Surveycited by 626
  4. 2021Documenting Large Webtext Corpora: A Case Study on the Colossal Clean Crawled Corpuscited by 481
  5. 2023NLP Evaluation in trouble: On the Need to Measure LLM Data Contamination for each Benchmarkcited by 339
  6. 2023From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Modelscited by 257
  7. 2023Large Language Model Alignment: A Surveycited by 241
  8. 2021What Will it Take to Fix Benchmarking in Natural Language Understanding?cited by 200
  9. 2023Evaluating Large Language Models: A Comprehensive Surveycited by 200
  10. 2023Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMscited by 191
  11. 2023The Capacity for Moral Self-Correction in Large Language Modelscited by 171
  12. 2021Neural Machine Translation for Low-Resource Languages: A Surveycited by 108
38 more
  1. 2024The Art of Saying No: Contextual Noncompliance in Language Modelscited by 100
  2. 2023Evaluating the Social Impact of Generative AI Systems in Systems and Societycited by 91
  3. 2023A Survey on Fairness in Large Language Modelscited by 91
  4. 2024Large Language Models are Geographically Biasedcited by 91
  5. 2021On Measures of Biases and Harms in NLPcited by 81
  6. 2025LLM Generated Persona is a Promise with a Catchcited by 77
  7. 2024BadChain: Backdoor Chain-of-Thought Prompting for Large Language Modelscited by 74
  8. 2023KoBBQ: Korean Bias Benchmark for Question Answeringcited by 71
  9. 2023Questioning the Survey Responses of Large Language Modelscited by 70
  10. 2023SeeGULL: A Stereotype Benchmark with Broad Geo-Cultural Coverage Leveraging Generative Modelscited by 66
  11. 2024TrojanRAG: Retrieval-Augmented Generation Can Be Backdoor Driver in Large Language Modelscited by 63
  12. 2024Fairness in Large Language Models: A Taxonomic Surveycited by 57
  13. 2023TrustGPT: A Benchmark for Trustworthy and Responsible Large Language Modelscited by 56
  14. 2022Perturbation Augmentation for Fairer NLPcited by 51
  15. 2022Characteristics of Harmful Text: Towards Rigorous Benchmarking of Language Modelscited by 49
  16. 2021Ethical-Advice Taker: Do Language Models Understand Natural Language Interventions?cited by 46
  17. 2021General-Purpose Question-Answering with Macawcited by 40
  18. 2024Social Bias Evaluation for Large Language Models Requires Prompt Variationscited by 39
  19. 2022Quantifying Social Biases Using Templates is Unreliablecited by 38
  20. 2025FairSteer: Inference Time Debiasing for LLMs with Dynamic Activation Steeringcited by 37
  21. 2024FairMT-Bench: Benchmarking Fairness for Multi-turn Dialogue in Conversational LLMscited by 36
  22. 2024KorNAT: LLM Alignment Benchmark for Korean Social Values and Common Knowledgecited by 33
  23. 2024MBBQ: A Dataset for Cross-Lingual Comparison of Stereotypes in Generative LLMscited by 32
  24. 2024Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compressioncited by 32
  25. 2022The Tail Wagging the Dog: Dataset Construction Biases of Social Bias Benchmarkscited by 31
  26. 2021Lawyers are Dishonest? Quantifying Representational Harms in Commonsense Knowledge Resourcescited by 30
  27. 2021Evaluating Debiasing Techniques for Intersectional Biasescited by 30
  28. 2025Actions Speak Louder than Words: Agent Decisions Reveal Implicit Biases in Language Modelscited by 30
  29. 2024LLMs are Biased Teachers: Evaluating LLM Bias in Personalized Educationcited by 28
  30. 2023Sociodemographic Bias in Language Models: A Survey and Forward Pathcited by 27
  31. 2024'Simulacrum of Stories': Examining Large Language Models as Qualitative Research Participantscited by 27
  32. 2024Does RAG Introduce Unfairness in LLMs? Evaluating Fairness in Retrieval-Augmented Generation Systemscited by 26
  33. 2023Building Socio-culturally Inclusive Stereotype Resources with Community Engagementcited by 25
  34. 2023KoSBi: A Dataset for Mitigating Social Bias Risks Towards Safer Large Language Model Applicationcited by 24
  35. 2022SODAPOP: Open-Ended Discovery of Social Biases in Social Commonsense Reasoning Modelscited by 22
  36. 2023CBBQ: A Chinese Bias Benchmark Dataset Curated with Human-AI Collaboration for Large Language Modelscited by 22
  37. 2024BiasAlert: A Plug-and-play Tool for Social Bias Detection in LLMscited by 21
  38. 2022A Survey for Efficient Open Domain Question Answeringcited by 20

50 most cited of 124 All 124 (JSON)