Research map / Large language models / Retrieval and knowledge
Retrospective Reader for Machine Reading Comprehension
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 sketchy and intensive reading with rear verification | ELECTRA · SQuAD2.0 test | F1: 91.4 -> 92.1 | Helped |
| Add sketchy and intensive reading with rear verification | ALBERT · NewsQA test | F1: 65.9 -> 66.8 | Helped |
| Add external front verification (ablation) | BERT · SQuAD2.0 dev | F1: 81.7 -> 82.1 | Helped |
| Add internal front verification with cross-entropy loss (ablation) | BERT · SQuAD2.0 dev | F1: 81.7 -> 82.0 | Helped |
| Add internal front verification with binary cross-entropy loss (ablation) | BERT · SQuAD2.0 dev | F1: 81.7 -> 81.8 | No clear effect |
| Add internal front verification with mean squared error loss (ablation) | BERT · SQuAD2.0 dev | F1: 81.7 -> 81.7 | No clear effect |
| Combine external and internal front verification as rear verification (ablation) | BERT · SQuAD2.0 dev | F1: 81.7 -> 82.5 | Helped |
| Combine external and internal front verification as rear verification (ablation) | ALBERT · SQuAD2.0 dev | F1: 90.2 -> 90.9 | Helped |
| Use retrospective reader instead of a two-model ensemble (ablation) | ALBERT · SQuAD2.0 dev | F1: 90.6 -> 90.9 | Helped |
| Add cross attention matching (ablation) | ALBERT · SQuAD2.0 dev | F1: 90.2 -> 90.3 | No clear effect |
| Add cross attention matching (ablation) | ALBERT · NewsQA dev | F1: 67.5 -> 66.3 | Hurt |
| Add matching attention (ablation) | ALBERT · SQuAD2.0 dev | F1: 90.2 -> 90.0 | Hurt |
Closest papers by meaning
- Read + Verify: Machine Reading Comprehension with Unanswerable Questions (AAAI 2019 · similarity 0.89)
- Assessing the Benchmarking Capacity of Machine Reading Comprehension Datasets (AAAI 2020 · similarity 0.88)
- From Cloze to Comprehension: Retrofitting Pre-trained Masked Language Models to Pre-trained Machine Reader (NeurIPS 2023 · similarity 0.88)
- MMM: Multi-Stage Multi-Task Learning for Multi-Choice Reading Comprehension (AAAI 2020 · similarity 0.87)
- NeurQuRI: Neural Question Requirement Inspector for Answerability Prediction in Machine Reading Comprehension (ICLR 2020 · similarity 0.87)
- PMRC: Prompt-Based Machine Reading Comprehension for Few-Shot Named Entity Recognition (AAAI 2024 · similarity 0.86)
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.