Research map / Vision-language and multimodal / Visual question answering
RINK: Reader-Inherited Evidence Reranker for Table-and-Text Open Domain 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 the reader-inherited RINK reranker combined with the instance-level reranker | Retriever-Reranker-Reader on OTT-QA · OTT-QA blind test | EM: 34.5 -> 35.5 | Helped |
| Use the reader-inherited RINK reranker combined with the instance-level reranker | Retriever-Reranker-Reader on OTT-QA · OTT-QA blind test | F1: 40.4 -> 41.5 | Helped |
| Add set-level RINK to the instance-level reranker | OTT-QA retrieval · OTT-QA dev | R@5: 77.4 -> 80.6 | Helped |
| Add set-level RINK to the instance-level reranker | OTT-QA retrieval · OTT-QA dev | R@1: 65.0 -> 66.9 | Helped |
| Pretrain the retriever encoder on augmented table-text data (ablation) | Retriever on OTT-QA · OTT-QA dev | R@5: 66.2 -> 70.2 | Helped |
| Pretrain the retriever encoder on augmented table-text data (ablation) | FiD QA system · OTT-QA blind test | EM: 28.5 -> 30.1 | Helped |
| Pretrain the retriever encoder on augmented table-text data (ablation) | RINK QA system · OTT-QA blind test | EM: 33.8 -> 35.5 | Helped |
| Increase RINK set sampling from K = 6 to K = 30 (ablation) | RINK with M = 15, α = 0 · OTT-QA dev | R@15: 82.8 -> 83.7 | Helped |
| Increase RINK set sampling from K = 6 to K = 30 (ablation) | RINK with M = 15, α = 0 · OTT-QA dev | EM: 33.4 -> 34.6 | Helped |
| Combine set-level RINK with the instance-level reranker instead of using RINK alone (ablation) | RINK with M = 15 · OTT-QA dev | R@15: 83.7 -> 85.5 | Helped |
| Set α to 0.7, combining the two rerankers (ablation) | RINK with M = 10 · OTT-QA dev | EM: 35.2 -> 36.7 | Helped |
Closest papers by meaning
- Evidence Aggregation for Answer Re-Ranking in Open-Domain Question Answering (ICLR 2018 · similarity 0.91)
- R3: Reinforced Ranker-Reader for Open-Domain Question Answering (AAAI 2018 · similarity 0.90)
- SuRe: Summarizing Retrievals using Answer Candidates for Open-domain QA of LLMs (ICLR 2024 · similarity 0.86)
- Autofocus Retrieval: An Effective Pipeline for Multi-Hop Question Answering With Semi-Structured Knowledge (NeurIPS 2026 · similarity 0.85)
- Making Retrieval-Augmented Language Models Robust to Irrelevant Context (ICLR 2024 · similarity 0.85)
- TableBench: A Comprehensive and Complex Benchmark for Table Question Answering (AAAI 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.