Research map / Vision-language and multimodal / Visual question answering
Convolutional Spatial Attention Model for Reading Comprehension with Multiple-Choice 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 |
|---|---|---|---|
| Add ELMo embeddings | CSA Model · RACE | accuracy: 48.4 -> 50.9 | Helped |
| Add ELMo embeddings | CSA Model · SemEval-2018 Task11 test | accuracy: 82.20 -> 83.27 | Helped |
| Use CSA instead of Dynamic Fusion Network | Single-model RACE system · RACE | accuracy: 47.4 -> 48.4 | Helped |
| Add ELMo embeddings and ensemble 9 CSA models | RACE system · RACE | accuracy: 51.2 -> 55.0 | Helped |
| Add element-wise trainable attention weights (ablation) | CSA Model · RACE | accuracy: 48.18 -> 48.52 | Helped |
| Add element-wise trainable attention weights (ablation) | CSA Model + ELMo · RACE | accuracy: 49.49 -> 50.89 | Helped |
| Add enriched candidate representations using question and passage interactions (ablation) | CSA Model · RACE | accuracy: 47.52 -> 48.52 | Helped |
| Add enriched candidate representations using question and passage interactions (ablation) | CSA Model + ELMo · RACE | accuracy: 49.78 -> 50.89 | Helped |
| Use convolutional spatial attention to summarize attention matrices (ablation) | CSA Model · RACE | accuracy: 47.30 -> 48.52 | Helped |
| Use convolutional spatial attention to summarize attention matrices (ablation) | CSA Model + ELMo · RACE | accuracy: 48.47 -> 50.89 | Helped |
| Apply CSA to RACE-H rather than relying on the previous best baseline | CSA Model · RACE-H | accuracy: 46.4 -> 47.3 | Helped |
| Apply CSA to RACE-M rather than relying on the previous best baseline | CSA Model · RACE-M | accuracy: 51.5 -> 51.0 | Hurt |
Closest papers by meaning
- Hierarchical Attention Flow for Multiple-Choice Reading Comprehension (AAAI 2018 · similarity 0.88)
- MMM: Multi-Stage Multi-Task Learning for Multi-Choice Reading Comprehension (AAAI 2020 · similarity 0.87)
- Hierarchical Question-Image Co-Attention for Visual Question Answering (NeurIPS 2016 · similarity 0.86)
- Generating Distractors for Reading Comprehension Questions from Real Examinations (AAAI 2019 · similarity 0.85)
- Multi-Matching Network for Multiple Choice Reading Comprehension (AAAI 2019 · similarity 0.85)
- FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering (NeurIPS 2025 · similarity 0.82)
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.