Counterfactual VQA: A Cause-Effect Look at Language Bias
Open accessThe DOI matches its Crossref record; 57 of the 57 references it cites were found in a registry or an index.
Abstract
VQA models may tend to rely on language bias as a shortcut and thus fail to sufficiently learn the multi-modal knowledge from both vision and language. Recent debiasing methods proposed to exclude the language prior during inference. However, they fail to disentangle the "good" language context and "bad" language bias from the whole. In this paper, we investigate how to mitigate language bias in VQA. Motivated by causal effects, we proposed a novel counterfactual inference framework, which enables us to capture the language bias as the direct causal effect of questions on answers and reduce the language bias by subtracting the direct language effect from the total causal effect. Experiments demonstrate that our proposed counterfactual inference framework 1) is general to various VQA backbones and fusion strategies, 2) achieves competitive performance on the language-bias sensitive VQA-CP dataset while performs robustly on the balanced VQA v2 dataset without any augmented data. The code is available at https://github.com/yuleiniu/cfvqa.
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
Distilling the Knowledge in a Neural Network
Hinton, G., Vinyals, O., & Dean, J. (2015) · arXiv · doi:10.48550/arxiv.1503.02531 · Cited by 14,204 · cited 1×
“(11)(13), which controls the sharpness of the distribution of Zq,v∗,k∗ like the softmax temperature [21].”
BackgroundMarginal Structural Models and Causal Inference in Epidemiology
Robins, J. M., Hernán, M. Á., & Brumback, B. (2000) · Epidemiology · doi:10.1097/00001648-200009000-00011 · Cited by 5,581
Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering
Anderson, P., He, X., Buehler, C., Teney, D., Johnson, M., Gould, S., & Zhang, L. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00636 · Cited by 4,948
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
The Book of Why: The New Science of Cause and Effect
Pearl, J., & Mackenzie, D. N. (2018) · Cited by 3,870
Causality: models, reasoning, and inference
(2000) · Choice Reviews Online · doi:10.5860/choice.38-1494a · Cited by 2,873
Bayesian inference for causal effects: The role of randomization
Rubin, D. B. (1978) · The Annals of Statistics · doi:10.1214/aos/1176344064 · Cited by 2,503
Robins, J. (1986) · Mathematical Modelling · doi:10.1016/0270-0255(86)90088-6 · Cited by 2,451 · cited 1×
“Causal effects reflect the comparisons between two potential outcomes of the same individual given two different treatments [36, 34].”
BackgroundMaking 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
Stacked Attention Networks for Image Question Answering
Yang, Z., He, X., Gao, J., Deng, L., & Smola, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.10 · Cited by 1,998
Cited by
Unbiased Scene Graph Generation From Biased Training
Tang, K., Niu, Y., Huang, J., Shi, J., & Zhang, H. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00377 · Cited by 755
Self-citationCausal Intervention for Weakly-Supervised Semantic Segmentation
Zhang, D., Zhang, H., Tang, J., Hua, X., & Sun, Q. (2020) · arXiv · doi:10.48550/arxiv.2009.12547 · Cited by 353
“Recently, there is a growing number of computer vision tasks that benefit from causality [34, 40, 51, 52, 56, 63, 65].”
BackgroundSelf-citationCross-Modal Causal Relational Reasoning for Event-Level Visual Question Answering
Liu, Y., Li, G., & Lin, L. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3284038 · Cited by 348
Multi-Modal Knowledge Graph Construction and Application: A Survey
Zhu, X., Li, Z., Wang, X., Jiang, X., Sun, P., Wang, X., Xiao, Y., & Yuan, N. J. (2024) · IEEE Transactions on Knowledge and Data Engineering · doi:10.1109/tkde.2022.3224228 · Cited by 291
Towards artificial general intelligence via a multimodal foundation model
Fei, N., Lu, Z., Gao, Y., Yang, G., Huo, Y., Wen, J., Lu, H., Song, R., Gao, X., Xiang, T., et al. (2022) · Nature Communications · doi:10.1038/s41467-022-30761-2 · Cited by 280
Self-citationModel-Agnostic Counterfactual Reasoning for Eliminating Popularity Bias in Recommender System
Wei, T., Feng, F., Chen, J., Wu, Z., Yi, J., & He, X. (2021) · Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining · doi:10.1145/3447548.3467289 · Cited by 274
Rahate, A., Walambe, R., Ramanna, S., & Kotecha, K. (2022) · Information Fusion · doi:10.1016/j.inffus.2021.12.003 · Cited by 214
“Making text input absent in VQA models, bias towards language modality is studied using a counterfactual approach [135].”
MethodA Review on Explainability in Multimodal Deep Neural Nets
Joshi, G., Walambe, R., & Kotecha, K. (2021) · IEEE Access · doi:10.1109/access.2021.3070212 · Cited by 204
“Such methods make multimodal frameworks more reliable, interpretable by allowing the models to provide reasonable predictions for the right reasons rather than looking for mere performance, looking at cause and effect aspects [185].”
BackgroundMethodMitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding
Leng, S., Zhang, H., Chen, G., Li, X., Lu, S., Miao, C., & Bing, L. (2023) · arXiv · doi:10.48550/arxiv.2311.16922 · Cited by 199
Self-citationMitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding
Leng, S., Zhang, H., Chen, G., Li, X., Lu, S., Miao, C., & Bing, L. (2024) · 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52733.2024.01316 · Cited by 199
Self-citationCausal Attention for Vision-Language Tasks
Yang, X., Zhang, H., Qi, G., & Cai, J. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00972 · Cited by 175
Self-citationLayout-based Causal Inference for Object Navigation
Zhang, S., Song, X., Li, W., Bai, Y., Yu, X., & Jiang, S. (2023) · 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr52729.2023.01039 · Cited by 41
CA-SpaceNet: Counterfactual analysis for 6D pose estimation in space
Wang, S., Wang, S., Jiao, B., Yang, D., Su, L., Zhai, P., Chen, C., & Zhang, L. (2022) · 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros47612.2022.9981172 · Cited by 24
“Recently, the computer vision community has paid more attention to the application of counterfactual analysis in many visual tasks such as longtailed visual recognition [27], action anticipation [28], scene graph generation (SGG) [29], and visual question answering (VQA) [30].”
BackgroundCausality-guided counterfactual debiasing for anomaly detection of cyber-physical systems
Tang, W., Liu, J., Zhou, Y., & Ding, Z. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2023.3326544 · Cited by 16
Cai, L., Fang, H., & Li, Z. (2023) · The Journal of Supercomputing · doi:10.1007/s11227-023-05195-2 · Cited by 8
“Inspired by causal effects, [16] captured the language prior as direct causal effect of questions on answers and reduced the language bias by subtracting the direct language effect from the total causal effect.”
Background
Metadata
- Authors
- Niu, Y., Tang, K., Zhang, H., Lu, Z., Hua, X.-S., & Wen, J.-R.
- Venue
- 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2021-06
- Pages
- 12695-12705
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W3177934633
- Cited by
- 418
- References
- 57
- FWCI
- 26.68
- Type
- conference paper
- Language
- en
- Open access
- green
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
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.