mlsanityData to 7 Oct 2026

RUBi: Reducing Unimodal Biases in Visual Question Answering

arXiv 1906.10169 ↗First posted 24 Jun 2019cs.CV cs.CL cs.LGcited by 183

Not on the map. Why?

Abstract (arXiv)

Visual Question Answering (VQA) is the task of answering questions about an image. Some VQA models often exploit unimodal biases to provide the correct answer without using the image information. As a result, they suffer from a huge drop in performance when evaluated on data outside their training set distribution. This critical issue makes them unsuitable for real-world settings. We propose RUBi, a new learning strategy to reduce biases in any VQA model. It reduces the importance of the most biased examples, i.e. examples that can be correctly classified without looking at the image. It implicitly forces the VQA model to use the two input modalities instead of relying on statistical regularities between the question and the answer. We leverage a question-only model that captures the language biases by identifying when these unwanted regularities are used. It prevents the base VQA model from learning them by influencing its predictions. This leads to dynamically adjusting the loss in order to compensate for biases. We validate our contributions by surpassing the current state-of-the-art results on VQA-CP v2. This dataset is specifically designed to assess the robustness of VQA models when exposed to different question biases at test time than what was seen during training. Our code is available: github.com/cdancette/rubi.bootstrap.pytorch

Key facts. Cited by 183 papers in the map (data through 7 Oct 2026). First posted on arXiv 24 Jun 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 52; 2022: 45; 2023: 32; 2024: 15; 2025: 21; 2026: 18.

183 papers.

Sort:

By area:

  1. 2021Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Conceptscited by 1,209
  2. 2022Discovering Invariant Rationales for Graph Neural Networkscited by 186
  3. 2021MultiBench: Multiscale Benchmarks for Multimodal Representation Learningcited by 143
  4. 2021Learning Debiased Representation via Disentangled Feature Augmentationcited by 116
  5. 2021EnD: Entangling and Disentangling deep representations for bias correctioncited by 105
  6. 2021Medical Visual Question Answering: A Surveycited by 93
  7. 2022Super-CLEVR: A Virtual Benchmark to Diagnose Domain Robustness in Visual Reasoningcited by 90
  8. 2021Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions with Superior OOD Generalizationcited by 83
  9. 2022Debiasing Graph Neural Networks via Learning Disentangled Causal Substructurecited by 83
  10. 2021Explainability of deep vision-based autonomous driving systems: Review and challengescited by 80
  11. 2021Causal Attention for Unbiased Visual Recognitioncited by 80
  12. 2021Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU Modelscited by 78
38 more
  1. 2021Beyond Question-Based Biases: Assessing Multimodal Shortcut Learning in Visual Question Answeringcited by 71
  2. 2021BiaSwap: Removing dataset bias with bias-tailored swapping augmentationcited by 70
  3. 2021Learning Task Informed Abstractionscited by 58
  4. 2021Which Shortcut Cues Will DNNs Choose? A Study from the Parameter-Space Perspectivecited by 56
  5. 2022Learning Debiased Classifier with Biased Committeecited by 51
  6. 2023Equivariant Similarity for Vision-Language Foundation Modelscited by 51
  7. 2023Large Language Models are Temporal and Causal Reasoners for Video Question Answeringcited by 51
  8. 2022Discover and Mitigate Unknown Biases with Debiasing Alternate Networkscited by 47
  9. 2021A Review on Explainability in Multimodal Deep Neural Netscited by 46
  10. 2021Unsupervised Learning of Debiased Representations with Pseudo-Attributescited by 44
  11. 2021Learning Fair Classifiers with Partially Annotated Group Labelscited by 40
  12. 2022Unbiased Supervised Contrastive Learningcited by 40
  13. 2022Gender and Racial Bias in Visual Question Answering Datasetscited by 39
  14. 2022SwapMix: Diagnosing and Regularizing the Over-Reliance on Visual Context in Visual Question Answeringcited by 38
  15. 2023VERITE: A Robust Benchmark for Multimodal Misinformation Detection Accounting for Unimodal Biascited by 37
  16. 2022MultiViz: Towards Visualizing and Understanding Multimodal Modelscited by 36
  17. 2021Multimodal Co-learning: Challenges, Applications with Datasets, Recent Advances and Future Directionscited by 35
  18. 2021Perceptual Score: What Data Modalities Does Your Model Perceive?cited by 35
  19. 2021Greedy Gradient Ensemble for Robust Visual Question Answeringcited by 34
  20. 2022DIME: Fine-grained Interpretations of Multimodal Models via Disentangled Local Explanationscited by 34
  21. 2022Language Prior Is Not the Only Shortcut: A Benchmark for Shortcut Learning in VQAcited by 30
  22. 2022SelecMix: Debiased Learning by Contradicting-pair Samplingcited by 29
  23. 2023Overcoming Simplicity Bias in Deep Networks using a Feature Sievecited by 29
  24. 2023Tackling Data Bias in MUSIC-AVQA: Crafting a Balanced Dataset for Unbiased Question-Answeringcited by 28
  25. 2023Survey of Social Bias in Vision-Language Modelscited by 26
  26. 2025Benchmark Designers Should "Train on the Test Set" to Expose Exploitable Non-Visual Shortcutscited by 26
  27. 2021Introspective Distillation for Robust Question Answeringcited by 25
  28. 2024ProtT3: Protein-to-Text Generation for Text-based Protein Understandingcited by 25
  29. 2023Improving Selective Visual Question Answering by Learning from Your Peerscited by 24
  30. 2023Synthetic Misinformers: Generating and Combating Multimodal Misinformationcited by 23
  31. 2023Robust Visual Question Answering: Datasets, Methods, and Future Challengescited by 23
  32. 2021Counterfactual Samples Synthesizing and Training for Robust Visual Question Answeringcited by 22
  33. 2022Mitigating Dataset Bias by Using Per-sample Gradientcited by 21
  34. 2022On Modality Bias Recognition and Reductioncited by 20
  35. 2023Debiased Fine-Tuning for Vision-language Models by Prompt Regularizationcited by 20
  36. 2022Revisiting the Importance of Amplifying Bias for Debiasingcited by 19
  37. 2022Rethinking Data Augmentation for Robust Visual Question Answeringcited by 18
  38. 2022Visual Perturbation-aware Collaborative Learning for Overcoming the Language Prior Problemcited by 18

50 most cited of 183 All 183 (JSON)