Research map / Vision-language and multimodal / Visual question answering
Open Question Answering over Tables and Text
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 |
|---|---|---|---|
| Replace iterative retrieval with fusion retrieval | Single-block reader on OTT-QA · OTT-QA Dev-Sparse | EM: 9.8 -> 14.3 | Helped |
| Replace the single-block reader with the cross-block reader | Iterative retriever on OTT-QA · OTT-QA Dev-Sparse | EM: 9.8 -> 17.1 | Helped |
| Combine fusion retrieval with cross-block reading | OTT-QA · OTT-QA Dev-Sparse | EM: 9.8 -> 27.7 | Helped |
| Use table-only retrieval instead of both tables and text (ablation) | Fusion-Retrieval + Cross-Block Reader · OTT-QA Test-Best | EM: 27.2 -> 4.4 | Hurt |
| Use text-only retrieval instead of both tables and text (ablation) | Fusion-Retrieval + Cross-Block Reader · OTT-QA Test-Best | EM: 27.2 -> 8.8 | Hurt |
| Use oracle table-passage links (ablation) | Fusion-Retrieval + Cross-Block Reader · OTT-QA Test-Best | EM: 27.2 -> 35.0 | Helped |
| Remove ICT pretraining and GPT-2 query augmentation (ablation) | Fusion-Retrieval + Cross-Block Reader · OTT-QA Dev-Dense | EM: 28.1 -> 24.6 | Hurt |
| Remove GPT-2 query augmentation (ablation) | Fusion-Retrieval + Cross-Block Reader · OTT-QA Dev-Dense | EM: 28.1 -> 22.1 | Hurt |
| Use fusion retrieval rather than iterative retrieval | Retriever, HITS@4K · OTT-QA sparse retrieval | HITS@4K: 35.8% to 48.1% | Helped |
| Use fusion retrieval rather than iterative retrieval | Retriever, HITS@4K · OTT-QA dense retrieval | HITS@4K: 27.2% to 52.4% | Helped |
| Add a third iterative retrieval hop after the second (ablation) | Iterative retriever · OTT-QA | HITS@4K | No clear effect |
| Use the proposed fusion retriever and cross-block reader | Open-domain multi-hop QA · HotpotQA Dev | EM: 50.4 vs. 52.7 | Mixed |
Closest papers by meaning
- Autofocus Retrieval: An Effective Pipeline for Multi-Hop Question Answering With Semi-Structured Knowledge (NeurIPS 2026 · similarity 0.87)
- TableBench: A Comprehensive and Complex Benchmark for Table Question Answering (AAAI 2025 · similarity 0.87)
- SuRe: Summarizing Retrievals using Answer Candidates for Open-domain QA of LLMs (ICLR 2024 · similarity 0.87)
- SPIQA: A Dataset for Multimodal Question Answering on Scientific Papers (NeurIPS 2024 · similarity 0.86)
- Text2Analysis: A Benchmark of Table Question Answering with Advanced Data Analysis and Unclear Queries (AAAI 2024 · similarity 0.85)
- Benchmarking Retrieval-Augmented Multimomal Generation for Document Question Answering (NeurIPS 2025 · 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.