Data to 7 Oct 2026

Don't Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biases

arXiv 1909.03683 ↗First posted 9 Sep 2019cs.CL cs.CV cs.LGcited by 279

Not on the map. Why?

Abstract (arXiv)

State-of-the-art models often make use of superficial patterns in the data that do not generalize well to out-of-domain or adversarial settings. For example, textual entailment models often learn that particular key words imply entailment, irrespective of context, and visual question answering models learn to predict prototypical answers, without considering evidence in the image. In this paper, we show that if we have prior knowledge of such biases, we can train a model to be more robust to domain shift. Our method has two stages: we (1) train a naive model that makes predictions exclusively based on dataset biases, and (2) train a robust model as part of an ensemble with the naive one in order to encourage it to focus on other patterns in the data that are more likely to generalize. Experiments on five datasets with out-of-domain test sets show significantly improved robustness in all settings, including a 12 point gain on a changing priors visual question answering dataset and a 9 point gain on an adversarial question answering test set.

Key facts. Cited by 279 papers in the map (data through 7 Oct 2026). First posted on arXiv 9 Sep 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 76; 2022: 74; 2023: 59; 2024: 30; 2025: 24; 2026: 16.

279 papers.

Sort:

By area:

2021Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Conceptscited by 1,209
2022DALL-Eval: Probing the Reasoning Skills and Social Biases of Text-to-Image Generation Modelscited by 230
2022MERLOT Reserve: Neural Script Knowledge through Vision and Language and Soundcited by 184
2022MultiHiertt: Numerical Reasoning over Multi Hierarchical Tabular and Textual Datacited by 120
2021Learning Debiased Representation via Disentangled Feature Augmentationcited by 116
2021Measure and Improve Robustness in NLP Models: A Surveycited by 107
2021Detoxifying Language Models Risks Marginalizing Minority Voicescited by 106
2021EnD: Entangling and Disentangling deep representations for bias correctioncited by 105
2023Revisiting Out-of-distribution Robustness in NLP: Benchmark, Analysis, and LLMs Evaluationscited by 96
2021Challenges in Automated Debiasing for Toxic Language Detectioncited by 95
2022On the Paradox of Learning to Reason from Datacited by 93
2021Competency Problems: On Finding and Removing Artifacts in Language Datacited by 91

38 more

2021Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions with Superior OOD Generalizationcited by 83
2022Shortcut Learning of Large Language Models in Natural Language Understandingcited by 83
2022Distilling Model Failures as Directions in Latent Spacecited by 82
2021Causal Attention for Unbiased Visual Recognitioncited by 80
2021Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU Modelscited by 78
2023Loose lips sink ships: Mitigating Length Bias in Reinforcement Learning from Human Feedbackcited by 78
2021Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Modelscited by 76
2021Beyond Question-Based Biases: Assessing Multimodal Shortcut Learning in Visual Question Answeringcited by 71
2023Discover and Cure: Concept-aware Mitigation of Spurious Correlationcited by 71
2021BiaSwap: Removing dataset bias with bias-tailored swapping augmentationcited by 70
2022Human-AI Collaboration via Conditional Delegation: A Case Study of Content Moderationcited by 66
2022Handling Bias in Toxic Speech Detection: A Surveycited by 60
2021Understanding the Effect of Out-of-distribution Examples and Interactive Explanations on Human-AI Decision Makingcited by 58
2021Learning Task Informed Abstractionscited by 58
2022Entropy-based Attention Regularization Frees Unintended Bias Mitigation from Listscited by 54
2022Generating Data to Mitigate Spurious Correlations in Natural Language Inference Datasetscited by 54
2022Learning Debiased Classifier with Biased Committeecited by 51
2021Fool Me Twice: Entailment from Wikipedia Gamificationcited by 50
2022Discover and Mitigate Unknown Biases with Debiasing Alternate Networkscited by 47
2021Can NLI Models Verify QA Systems' Predictions?cited by 46
2021Unsupervised Learning of Debiased Representations with Pseudo-Attributescited by 44
2021An Investigation of the (In)effectiveness of Counterfactually Augmented Datacited by 42
2022Unbiased Supervised Contrastive Learningcited by 40
2022Gender and Racial Bias in Visual Question Answering Datasetscited by 39
2022SwapMix: Diagnosing and Regularizing the Over-Reliance on Visual Context in Visual Question Answeringcited by 38
2024Core Knowledge Deficits in Multi-Modal Language Modelscited by 38
2025PerturboLLaVA: Reducing Multimodal Hallucinations with Perturbative Visual Trainingcited by 37
2021Perceptual Score: What Data Modalities Does Your Model Perceive?cited by 35
2021Greedy Gradient Ensemble for Robust Visual Question Answeringcited by 34
2021Select, Substitute, Search: A New Benchmark for Knowledge-Augmented Visual Question Answeringcited by 33
2024MM-SpuBench: Towards Better Understanding of Spurious Biases in Multimodal LLMscited by 33
2021Learning Bias-Invariant Representation by Cross-Sample Mutual Information Minimizationcited by 32
2024On the Challenges and Opportunities in Generative AIcited by 31
2022Bias Mimicking: A Simple Sampling Approach for Bias Mitigationcited by 30
2022Language Prior Is Not the Only Shortcut: A Benchmark for Shortcut Learning in VQAcited by 30
2021Supervising Model Attention with Human Explanations for Robust Natural Language Inferencecited by 29
2022DISCO: Distilling Counterfactuals with Large Language Modelscited by 29
2023Overcoming Simplicity Bias in Deep Networks using a Feature Sievecited by 29

50 most cited of 279 All 279 (JSON)