Research map / Vision-language and multimodal / Visual question answering
Actively Seeking and Learning From Live Data
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 |
|---|---|---|---|
| Adapt the VQA model using retrieved training QAs | VQA model · VQA-CP v2 Test | accuracy: 40.71 -> 46.00 | Helped |
| Use COCO captions as support data instead of QA support | VQA model · VQA-CP v2 Test | accuracy: 39.84 vs 40.71 | Mixed |
| Train caption-supported adaptation only on Other questions | VQA model with COCO caption support · VQA-CP v2 Test, Other | accuracy: 42.93 -> 43.95 | Helped |
| Provide QAs from the test distribution as leave-one-out support | VQA model trained on VQA-CP · VQA-CP v2 Test | accuracy: 46.00 -> 52.09 | Helped |
| Use uniform support retrieval rather than no adaptation (ablation) | VQA model · VQA-CP v2 Other, Test, QA support | accuracy: 31.09 -> 31.33 | Helped |
| Use uniform support retrieval rather than no adaptation (ablation) | VQA model · VQA-CP v2 Other, Test, COCO caption support | accuracy: 31.09 -> 34.00 | Helped |
| Retrieve support using common question words (ablation) | VQA model with QA support · VQA-CP v2 Other, Test | accuracy: 31.33 vs 31.79 | Helped |
| Retrieve support using common question words (ablation) | VQA model with COCO caption support · VQA-CP v2 Other, Test | accuracy: 34.00 vs 29.18 | Hurt |
| Retrieve support containing a top-5 baseline answer word (ablation) | VQA model with QA support · VQA-CP v2 Other, Test | accuracy: 31.33 vs 31.76 | Helped |
| Use image-feature similarity for support retrieval (ablation) | VQA model with QA support · VQA-CP v2 Other, Test | accuracy: 31.33 vs 31.68 | Helped |
| Combine answer-word and image-similarity retrieval (ablation) | VQA model with QA support · VQA-CP v2 Other, Test | accuracy: 31.33 vs 31.09 | Hurt |
| Combine word-overlap, answer-word, and image-similarity retrieval (ablation) | VQA model with QA support · VQA-CP v2 Other, Test | accuracy: 31.33 vs 34.25 | Helped |
Closest papers by meaning
- Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering (CVPR 2018 · similarity 0.92)
- Answer-Type Prediction for Visual Question Answering (CVPR 2016 · similarity 0.91)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.91)
- Learning Answer Embeddings for Visual Question Answering (CVPR 2018 · similarity 0.90)
- Making the v in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering (CVPR 2017 · similarity 0.90)
- Cross-Dataset Adaptation for Visual Question Answering (CVPR 2018 · 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.