mlsanityData to 7 Oct 2026

Towards Debiasing NLU Models from Unknown Biases

arXiv 2009.12303 ↗First posted 25 Sep 2020cs.CL cs.AI cs.LGcited by 116

Not on the map. Why?

Abstract (arXiv)

NLU models often exploit biases to achieve high dataset-specific performance without properly learning the intended task. Recently proposed debiasing methods are shown to be effective in mitigating this tendency. However, these methods rely on a major assumption that the types of bias should be known a-priori, which limits their application to many NLU tasks and datasets. In this work, we present the first step to bridge this gap by introducing a self-debiasing framework that prevents models from mainly utilizing biases without knowing them in advance. The proposed framework is general and complementary to the existing debiasing methods. We show that it allows these existing methods to retain the improvement on the challenge datasets (i.e., sets of examples designed to expose models' reliance on biases) without specifically targeting certain biases. Furthermore, the evaluation suggests that applying the framework results in improved overall robustness.

Key facts. Cited by 116 papers in the map (data through 7 Oct 2026). First posted on arXiv 25 Sep 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 31; 2022: 31; 2023: 28; 2024: 10; 2025: 11; 2026: 5.

116 papers.

Sort:

By area:

  1. 2023Bias and Fairness in Large Language Models: A Surveycited by 626
  2. 2021Just Train Twice: Improving Group Robustness without Training Group Informationcited by 522
  3. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 412
  4. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 379
  5. 2021A Survey on Automated Fact-Checkingcited by 376
  6. 2021Get Your Vitamin C! Robust Fact Verification with Contrastive Evidencecited by 222
  7. 2021Measure and Improve Robustness in NLP Models: A Surveycited by 107
  8. 2023A Survey on Fairness in Large Language Modelscited by 91
  9. 2021Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions with Superior OOD Generalizationcited by 83
  10. 2022Shortcut Learning of Large Language Models in Natural Language Understandingcited by 83
  11. 2022Distilling Model Failures as Directions in Latent Spacecited by 82
  12. 2021Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU Modelscited by 78
38 more
  1. 2023Discover and Cure: Concept-aware Mitigation of Spurious Correlationcited by 71
  2. 2021Generalization in NLI: Ways (Not) To Go Beyond Simple Heuristicscited by 67
  3. 2021Fool Me Twice: Entailment from Wikipedia Gamificationcited by 50
  4. 2022Stretching Sentence-pair NLI Models to Reason over Long Documents and Clusterscited by 44
  5. 2021An Investigation of the (In)effectiveness of Counterfactually Augmented Datacited by 42
  6. 2025Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuningcited by 34
  7. 2024A Survey of Small Language Modelscited by 31
  8. 2022On Guiding Visual Attention with Language Specificationcited by 30
  9. 2021Modeling the Second Player in Distributionally Robust Optimizationcited by 29
  10. 2021Supervising Model Attention with Human Explanations for Robust Natural Language Inferencecited by 29
  11. 2023From Shortcuts to Triggers: Backdoor Defense with Denoised PoEcited by 26
  12. 2021Training Dynamics for Text Summarization Modelscited by 25
  13. 2021BARACK: Partially Supervised Group Robustness With Guaranteescited by 25
  14. 2021End-to-End Self-Debiasing Framework for Robust NLU Trainingcited by 24
  15. 2022A Robust Bias Mitigation Procedure Based on the Stereotype Content Modelcited by 24
  16. 2021Avoiding Inference Heuristics in Few-shot Prompt-based Finetuningcited by 23
  17. 2022BLIND: Bias Removal With No Demographicscited by 22
  18. 2023Weakly Supervised Detection of Hallucinations in LLM Activationscited by 20
  19. 2021Context-aware Adversarial Training for Name Regularity Bias in Named Entity Recognitioncited by 19
  20. 2021Measuring and Improving Model-Moderator Collaboration using Uncertainty Estimationcited by 18
  21. 2021RAIL-KD: RAndom Intermediate Layer Mapping for Knowledge Distillationcited by 18
  22. 2022Saving Dense Retriever from Shortcut Dependency in Conversational Searchcited by 18
  23. 2021Debiasing Methods in Natural Language Understanding Make Bias More Accessiblecited by 17
  24. 2022Logical Reasoning with Span-Level Predictions for Interpretable and Robust NLI Modelscited by 17
  25. 2021Predict then Interpolate: A Simple Algorithm to Learn Stable Classifierscited by 16
  26. 2021Uncertainty Calibration for Ensemble-Based Debiasing Methodscited by 16
  27. 2022Learning to Split for Automatic Bias Detectioncited by 16
  28. 2022Debiasing Masks: A New Framework for Shortcut Mitigation in NLUcited by 16
  29. 2021Are Bias Mitigation Techniques for Deep Learning Effective?cited by 14
  30. 2022Less Learn Shortcut: Analyzing and Mitigating Learning of Spurious Feature-Label Correlationcited by 13
  31. 2024Take Care of Your Prompt Bias! Investigating and Mitigating Prompt Bias in Factual Knowledge Extractioncited by 13
  32. 2021Robustness Challenges in Model Distillation and Pruning for Natural Language Understandingcited by 12
  33. 2022Audio-based AI classifiers show no evidence of improved COVID-19 screening over simple symptoms checkerscited by 12
  34. 2024Complexity Matters: Dynamics of Feature Learning in the Presence of Spurious Correlationscited by 12
  35. 2022OccamNets: Mitigating Dataset Bias by Favoring Simpler Hypothesescited by 11
  36. 2022Kernel-Whitening: Overcome Dataset Bias with Isotropic Sentence Embeddingcited by 11
  37. 2023Understanding and Mitigating Spurious Correlations in Text Classification with Neighborhood Analysiscited by 11
  38. 2023Tackling Bias in Pre-trained Language Models: Current Trends and Under-represented Societiescited by 11

50 most cited of 116 All 116 (JSON)