Research map / Vision-language and multimodal / Visual question answering
A Unified Model for Document-Based Question Answering Based on Human-Like Reading Strategy
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 |
|---|---|---|---|
| concatenate title encoding into question encoding (ablation) | CA-Network · WikiQA test | MAP: 0.7433 -> 0.7541 | Helped |
| add title to question by element-wise addition (ablation) | CA-Network · WikiQA test | MAP: 0.7433 -> 0.7429 | No clear effect |
| combine title and question by element-wise multiplication (ablation) | CA-Network · WikiQA test | MAP: 0.7433 -> 0.7056 | Hurt |
| attention-weighted title vector as question representation (ablation) | CA-Network · WikiQA test | MAP: 0.7433 -> 0.7102 | Hurt |
| add attended title to original question (qat-add-q) (ablation) | CA-Network · WikiQA test | MAP: 0.7433 -> 0.7424 | No clear effect |
| use LDA topic instead of title as summary (ablation) | q-t · WikiQA test | MAP: 0.7541 -> 0.7441 | Hurt |
| use LSA topic instead of title as summary (ablation) | q-t · WikiQA test | MAP: 0.7541 -> 0.7247 | Hurt |
| use first sentence as summary instead of title (ablation) | q-t · WikiQA test | MAP: 0.7541 -> 0.7193 | Hurt |
| use last sentence as summary instead of title (ablation) | q-t · WikiQA test | MAP: 0.7541 -> 0.7304 | Hurt |
| add LDA-topic summary to question encoding | CA-Network · NLPCC2016 Chinese DBQA test | MAP: 0.8073 -> 0.8443 | Helped |
Closest papers by meaning
- A Deep Cascade Model for Multi-Document Reading Comprehension (AAAI 2019 · similarity 0.88)
- Medical Exam Question Answering with Large-scale Reading Comprehension (AAAI 2018 · similarity 0.87)
- NovelQA: Benchmarking Question Answering on Documents Exceeding 200K Tokens (ICLR 2025 · similarity 0.86)
- An Empirical Study of Content Understanding in Conversational Question Answering (AAAI 2020 · similarity 0.86)
- Visual Question Answering as Reading Comprehension (CVPR 2019 · similarity 0.86)
- GRAM: Global Reasoning for Multi-Page VQA (CVPR 2024 · similarity 0.85)
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.