Research map / Vision-language and multimodal / Visual question answering
VisFIS: Visual Feature Importance Supervision with Right-for-the-Right-Reason Objectives
Zhuofan Ying, Peter Hase, Mohit Bansal
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add VisFIS human feature-importance supervision | UpDn · CLEVR-XAI OOD | accuracy: 39.08 -> 43.78 | Improved |
| add VisFIS human feature-importance supervision | UpDn · CLEVR-XAI ID | accuracy: 71.72 -> 72.82 | Improved |
| add VisFIS on GQA-101k | UpDn · GQA-101k OOD | accuracy: 31.80 -> 34.88 | Improved |
| add VisFIS on VQA-HAT | UpDn · VQA-HAT OOD | accuracy: 28.76 -> 31.21 | Improved |
| add VisFIS | LXMERT · CLEVR-XAI OOD | accuracy: 73.76 -> 74.83 | Improved |
| replace human explanations with random explanations in VisFIS (ablation) | VisFIS · CLEVR-XAI OOD | accuracy: 43.78 -> 33.28 | Hurt |
| add Align objective alone (ablation) | UpDn · CLEVR-XAI OOD | accuracy: 36.80 -> 41.61 | Improved |
| add Suff-Human objective alone (ablation) | UpDn · CLEVR-XAI OOD | accuracy: 36.80 -> 40.91 | Improved |
| add Inv-FI objective alone (ablation) | UpDn · CLEVR-XAI OOD | accuracy: 36.80 -> 38.88 | Improved |
| add Inv-DA objective instead of Inv-FI (ablation) | UpDn · CLEVR-XAI OOD | accuracy: 36.80 -> 35.91 | Hurt |
| combine all four objectives in VisFIS versus Suff-Human + Align (ablation) | UpDn · CLEVR-XAI OOD | accuracy: 41.63 -> 43.78 | Improved |
| random supervision on the Suff-Human term (ablation) | VisFIS · CLEVR-XAI OOD | accuracy: 43.78 -> 36.70 | Hurt |
Closest papers by meaning
- OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge (CVPR 2019 · similarity 0.89)
- Making the v in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering (CVPR 2017 · similarity 0.89)
- Exploring the Effectiveness of Object-Centric Representations in Visual Question Answering: Comparative Insights with Foundation Models (ICLR 2025 · similarity 0.89)
- Human Uncertainty-Aware Data Selection and Automatic Labeling in Visual Question Answering (ICLR 2026 · similarity 0.89)
- VisRes Bench: On Evaluating the Visual Reasoning Capabilities of VLMs (CVPR 2026 · similarity 0.89)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
