Research map / Vision-language and multimodal / Visual question answering
Found a Reason for me? Weakly-supervised Grounded Visual Question Answering using Capsules
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 query-based soft-masked capsules to MAC | MAC · CLEVR-Answers val | IOU F1: 19.73 -> 36.07 | Helped |
| Add query-based soft-masked capsules to SNMN | SNMN · CLEVR-Answers val | IOU F1: 42.11 -> 51.18 | Helped |
| Add query-based soft-masked capsules to MAC | MAC · GQA val, full answer grounding | Overlap F1: 26.15 -> 50.06 | Helped |
| Add query-based soft-masked capsules to MAC | MAC · GQA val | accuracy: 57.09 -> 55.13 | Hurt |
| Replace capsules with query-masked convolutional features (ablation) | SNMN-Caps · CLEVR-Answers val | IOU F1: 50.42 -> 45.49 | Hurt |
| Replace soft capsule masking with one-hot hard masking (ablation) | SNMN-Caps · CLEVR-Answers val | accuracy: 95.79 -> 88.48 | Hurt |
| Share masking-layer weights across reasoning modules (ablation) | SNMN-Caps · CLEVR-Answers val | IOU F1: 50.42 -> 48.87 | Hurt |
| Use 8 capsules instead of 16 (ablation) | SNMN-Caps · CLEVR-Answers val | IOU F1: 50.42 -> 44.67 | Hurt |
| Use 24 capsules instead of 16 (ablation) | SNMN-Caps · CLEVR-Answers val | accuracy: 95.79 -> 96.66 | Helped |
| Increase the opacity parameter for uniform attention regions (ablation) | SNMN-Caps · CLEVR-Answers grounding | F1 | Helped |
Closest papers by meaning
- Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering (CVPR 2018 · similarity 0.91)
- Dynamic Capsule Attention for Visual Question Answering (AAAI 2019 · similarity 0.90)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.89)
- Can I Trust Your Answer? Visually Grounded Video Question Answering (CVPR 2024 · similarity 0.89)
- FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering (NeurIPS 2025 · similarity 0.89)
- VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning (CVPR 2024 · 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.