Research map / Vision-language and multimodal / Visual question answering
Learning Conditioned Graph Structures for Interpretable Visual Question Answering
Read the paperCheck 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 |
|---|---|---|---|
| Use a question-conditioned learned graph with spatial graph convolutions instead of question-to-image attention | VQA model · VQA 2.0 standard test set | All accuracy: 61.90 -> 66.18 | Helped |
| Use a learned, attention-weighted graph instead of a k-nearest-neighbor graph | VQA model · VQA 2.0 standard test set | All accuracy: 61.00 -> 66.18 | Helped |
| Use a question-conditioned learned graph with spatial graph convolutions instead of question-to-image attention | VQA model · VQA 2.0 standard test set | Num. accuracy: 42.48 -> 47.13 | Helped |
| Use a question-conditioned learned graph with spatial graph convolutions instead of question-to-image attention | VQA model · VQA 2.0 standard test set | Y/N accuracy: 79.87 -> 82.91 | Helped |
| Use a question-conditioned learned graph with spatial graph convolutions instead of question-to-image attention | VQA model · VQA 2.0 standard test set | Other accuracy: 50.95 -> 56.22 | Helped |
| Reduce the neighborhood size below 16 (ablation) | Ours · VQA 2.0 | classification accuracy | Hurt |
| Increase the neighborhood size above 16 (ablation) | Ours · VQA 2.0 | classification accuracy | No clear effect |
| Vary the number of Gaussian kernels away from 8 (ablation) | Ours · VQA 2.0 | classification accuracy | Hurt |
| Use polar rather than Cartesian pseudo-coordinates (ablation) | Ours · VQA 2.0 | classification accuracy | Helped |
Closest papers by meaning
- Graph-Structured Representations for Visual Question Answering (CVPR 2017 · similarity 0.93)
- Visual Question Answering with Question Representation Update (QRU) (NeurIPS 2016 · similarity 0.92)
- Learning Answer Embeddings for Visual Question Answering (CVPR 2018 · similarity 0.91)
- Dynamic Key-Value Memory Enhanced Multi-Step Graph Reasoning for Knowledge-Based Visual Question Answering (AAAI 2022 · similarity 0.91)
- Bridge To Answer: Structure-Aware Graph Interaction Network for Video Question Answering (CVPR 2021 · similarity 0.91)
- Multimodal Commonsense Knowledge Distillation for Visual Question Answering (Student Abstract) (AAAI 2025 · similarity 0.90)
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.