Research map / Vision-language and multimodal / Visual question answering
Don't Just Assume; Look and Answer: Overcoming Priors for 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 GVQA's grounded, disentangled architecture rather than SAN | VQA-CP v1 · VQA-CP v1 | overall accuracy: 26.88 -> 39.23 | Helped |
| Use GVQA's grounded, disentangled architecture rather than SAN | VQA-CP v2 · VQA-CP v2 | overall accuracy: 24.96 -> 31.30 | Helped |
| Use GVQA instead of MCB | VQA-CP v1 · VQA-CP v1 | overall accuracy: 34.39 -> 39.23 | Helped |
| Use GVQA rather than SAN | VQA v1 · VQA v1 | overall accuracy: 55.86 -> 51.12 | Hurt |
| Use GVQA rather than SAN | VQA v2 · VQA v2 | overall accuracy: 52.02 -> 48.24 | Hurt |
| Use Q_main rather than Q_full in VCC (ablation) | GVQA · VQA-CP v1 | overall accuracy: 33.55 -> 39.23 | Helped |
| Replace CE with an end-to-end LSTM (ablation) | GVQA yes/no pipeline · VQA-CP v1 | overall accuracy: 39.23 -> 27.28; drop of 11.95% | Hurt |
| Replace ACP with an end-to-end LSTM (ablation) | GVQA · VQA-CP v1 | overall accuracy: 39.23 -> 39.40 | No clear effect |
| Remove VCC loss (ablation) | GVQA · VQA-CP v1 | overall accuracy: +1.72% | Helped |
| Add ACP rather than replace it with LSTM, when VCC loss is absent (ablation) | GVQA without VCC loss · VQA-CP v1 | overall accuracy: 38.86 -> 40.95 | Helped |
| Use Q_main rather than Q_full (ablation) | SAN · VQA-CP v2 | overall accuracy: 24.96 -> 26.32 | Helped |
| Use 25 rather than 50 answer clusters (ablation) | GVQA ACP · VQA-CP v2 | overall accuracy: 1.05% drop | Hurt |
Closest papers by meaning
- Making the v in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering (CVPR 2017 · similarity 0.95)
- Tips and Tricks for Visual Question Answering: Learnings From the 2017 Challenge (CVPR 2018 · similarity 0.94)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.94)
- Roses Are Red, Violets Are Blue... but Should VQA Expect Them To? (CVPR 2021 · similarity 0.94)
- Answer Them All! Toward Universal Visual Question Answering Models (CVPR 2019 · similarity 0.93)
- Self-Critical Reasoning for Robust Visual Question Answering (NeurIPS 2019 · similarity 0.93)
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.