Research map / Vision-language and multimodal / Visual question answering
Learning by Asking Questions
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 LBA-generated training examples instead of the CLEVR train set | CNN+LSTM+SA on CLEVR · CLEVR val | accuracy: 77.2 -> 78.2 | Helped |
| condition the question generator on question type (ablation) | LBA · CLEVR val | accuracy: 54.7 -> 63.4 | Helped |
| add a learned relevance model to filter proposed questions (ablation) | LBA with image- and question-type-conditioned generator · CLEVR val | accuracy: 63.4 -> 78.2 | Helped |
| use a perfect relevance model instead of the learned relevance model (ablation) | LBA · CLEVR val | accuracy: 78.2 -> 81.1 | Helped |
| increase generator sampling temperature from 0.3 to 1.3 (ablation) | LBA with relevance filtering · CLEVR val | accuracy: 70.1 -> 78.2 | Helped |
| use the curriculum question-selection policy rather than random selection (ablation) | LBA · CLEVR | accuracy | Helped |
| select questions with entropy-based policies rather than random selection (ablation) | LBA · CLEVR | accuracy | Hurt |
| reduce the bootstrap set size from 70k to 20k examples (ablation) | LBA · CLEVR val | accuracy: 79.1 -> 76.2 | Hurt |
| use offline VQA training on bootstrap and collected examples (ablation) | LBA · CLEVR | accuracy | Helped |
Closest papers by meaning
- Knowledge Acquisition for Visual Question Answering via Iterative Querying (CVPR 2017 · similarity 0.91)
- Roses Are Red, Violets Are Blue... but Should VQA Expect Them To? (CVPR 2021 · similarity 0.90)
- VQ-VA World: Towards High-Quality Visual Question-Visual Answering (CVPR 2026 · similarity 0.89)
- OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge (CVPR 2019 · similarity 0.89)
- VQA With No Questions-Answers Training (CVPR 2020 · 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.