Adversarial Regularization for Visual Question Answering: Strengths, Shortcomings, and Side Effects
Open accessThe DOI matches its Crossref record; 10 of the 10 references it cites were found in a registry or an index.
Abstract
Visual question answering (VQA) models have been shown to over-rely on linguistic biases in VQA datasets, answering questions "blindly" without considering visual context.Adversarial regularization (AdvReg) aims to address this issue via an adversary subnetwork that encourages the main model to learn a bias-free representation of the question.In this work, we investigate the strengths and shortcomings of AdvReg with the goal of better understanding how it affects inference in VQA models.Despite achieving a new stateof-the-art on VQA-CP, we find that AdvReg yields several undesirable side-effects, including unstable gradients and sharply reduced performance on in-domain examples.We demonstrate that gradual introduction of regularization during training helps to alleviate, but not completely solve, these issues.Through error analyses, we observe that AdvReg improves generalization to binary questions, but impairs performance on questions with heterogeneous answer distributions.Qualitatively, we also find that regularized models tend to over-rely on visual features, while ignoring important linguistic cues in the question.Our results suggest that AdvReg requires further refinement before it can be considered a viable bias mitigation technique for VQA.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
VQA: Visual Question Answering
Antol, S., Agrawal, A., Lu, J., Mitchell, M., Batra, D., Zitnick, C. L., & Parikh, D. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.279 · Cited by 4,487
Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
Goyal, Y., Khot, T., Summers-Stay, D., Batra, D., & Parikh, D. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.670 · Cited by 2,287
CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning
Johnson, J., Hariharan, B., van der Maaten, L., Fei-Fei, L., Zitnick, C. L., & Girshick, R. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.215 · Cited by 1,781
Annotation Artifacts in Natural Language Inference Data
Gururangan, S., Swayamdipta, S., Levy, O., Schwartz, R., Bowman, S., & Smith, N. A. (2018) · Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers) · doi:10.18653/v1/n18-2017 · Cited by 907
Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints
Zhao, J., Wang, T., Yatskar, M., Ordonez, V., & Chang, K.-W. (2017) · Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d17-1323 · Cited by 829 · cited 1×
“For instance, the Commmon Objects in Context (COCO) image dataset (Lin et al., 2014), which is used in VQA, has been shown to contain prominent gender biases (Zhao et al., 2017; Hendricks et al., 2018).”
BackgroundVQA: Visual Question Answering: www.visualqa.org
Agrawal, A., Lu, J., Antol, S., Mitchell, M., Zitnick, C. L., Parikh, D., & Batra, D. (2017) · International Journal of Computer Vision · doi:10.1007/s11263-016-0966-6 · Cited by 570
Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
Goyal, Y., Khot, T., Agrawal, A., Summers-Stay, D., Batra, D., & Parikh, D. (2019) · International Journal of Computer Vision · doi:10.1007/s11263-018-1116-0 · Cited by 487 · cited 8×
“?”) remains skewed towards a handful of top answers (Goyal et al., 2017).”
BackgroundMethodInfluentialWomen Also Snowboard: Overcoming Bias in Captioning Models
Hendricks, L. A., Burns, K., Saenko, K., Darrell, T., & Rohrbach, A. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01219-9_47 · Cited by 404
Analyzing the Behavior of Visual Question Answering Models
Agrawal, A., Batra, D., & Parikh, D. (2016) · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d16-1203 · Cited by 297 · cited 2×
“Biases in VQA datasets A growing body of work points to the existence of biases in popular VQA datasets (Agrawal et al., 2016; Zhang et al., 2016; Jabri et al., 2016; Goyal et al., 2017; Johnson et al., 2017; Chao et al., 2018; Agrawal et al., 2018; Thomason et al., 2019).”
BackgroundCognitive Biases in Crowdsourcing
Eickhoff, C. (2018) · Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining · doi:10.1145/3159652.3159654 · Cited by 144 · cited 1×
“any benchmark that involves crowdsourced data is likely to encode certain cognitive and/or social biases (van Miltenburg, 2016; Misra et al., 2016; Eickhoff, 2018).”
Background
Cited by
Don’t Take the Easy Way Out: Ensemble Based Methods for Avoiding Known Dataset Biases
Clark, C., Yatskar, M., & Zettlemoyer, L. (2019) · Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) · doi:10.18653/v1/d19-1418 · Cited by 381
“Instead we follow prior work (Grand and Belinkov, 2019; Ramakr-ishnan et al., 2018) and perform model selection on the test set.”
BackgroundMethodCounterfactual Samples Synthesizing for Robust Visual Question Answering
Chen, L., Yan, X., Xiao, J., Zhang, H., Pu, S., & Zhuang, Y. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.01081 · Cited by 341
Koçak, B., Ponsiglione, A., Stanzione, A., Bluethgen, C., Santinha, J., Ugga, L., Huisman, M., Klontzas, M. E., Cannella, R., & Cuocolo, R. (2024) · Diagnostic and Interventional Radiology · doi:10.4274/dir.2024.242854 · Cited by 168
Don’t Take the Premise for Granted: Mitigating Artifacts in Natural Language Inference
Belinkov, Y., Poliak, A., Shieber, S., Van Durme, B., & Rush, A. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1084 · Cited by 73
Self-citationRemoving Bias in Multi-modal Classifiers: Regularization by Maximizing Functional Entropies
Gat, I., Schwartz, I., Schwing, A., & Hazan, T. (2020) · arXiv · doi:10.48550/arxiv.2010.10802 · Cited by 63
“The authors of [47, 48] raise the concern that new regularization methods mainly boost the performance of yes/no questions.”
Background
Metadata
- Authors
- Grand, G., & Belinkov, Y.
- Venue
- Proceedings of the Second Workshop on Shortcomings in Vision and Language
- Published
- 2019
- Pages
- 1-13
- OpenAlex
- W2952111037
- Cited by
- 76
- References
- 10
- FWCI
- 4.30
- Type
- conference paper
- Language
- en
- Open access
- gold
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Adversarial Regularization for Visual Question Answering: Strengths, Shortcomings, and Side Effects. https://app.cochat.ai/papers/10.18653/v1/w19-1801
- The paper
- Grand, G., & Belinkov, Y. (2019). Adversarial Regularization for Visual Question Answering: Strengths, Shortcomings, and Side Effects. Proceedings of the Second Workshop on Shortcomings in Vision and Language. https://doi.org/10.18653/v1/w19-1801
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.