Research map / Architectures and efficiency / Attention and transformers
Bidirectional Attention Flow for Machine 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 |
|---|---|---|---|
| Use BiDAF as a single model rather than the strongest single-model baseline | SQuAD test · SQuAD test | EM: 68.4 vs 68.0 | Mixed |
| Use BiDAF as an ensemble rather than the strongest listed ensemble baseline | SQuAD test · SQuAD test | F1: 80.4 -> 81.1 | Helped |
| Use BiDAF rather than prior systems | CNN · CNN test | accuracy: 74.7 -> 76.9 | Helped |
| Use BiDAF rather than prior systems | DailyMail · DailyMail test | accuracy: 76.6 -> 79.6 | Helped |
| Add character embeddings (ablation) | BiDAF · SQuAD dev | EM: 65.0 -> 67.7 | Helped |
| Add word embeddings (ablation) | BiDAF · SQuAD dev | EM: 55.5 -> 67.7 | Helped |
| Add C2Q attention (ablation) | BiDAF · SQuAD dev | EM: 57.2 -> 67.7 | Helped |
| Add Q2C attention (ablation) | BiDAF · SQuAD dev | EM: 63.6 -> 67.7 | Helped |
| Use static, precomputed attention rather than dynamic attention (ablation) | BiDAF · SQuAD dev | EM: 63.5 -> 67.7 | Helped |
| Use dot-product similarity (ablation) | BiDAF · SQuAD dev | EM: 68.0 -> 65.5 | Hurt |
| Use linear similarity (ablation) | BiDAF · SQuAD dev | EM: 68.0 -> 59.5 | Hurt |
| Use bilinear similarity (ablation) | BiDAF · SQuAD dev | EM: 68.0 -> 61.6 | Hurt |
Closest papers by meaning
- Answering Complex Queries in Knowledge Graphs with Bidirectional Sequence Encoders (AAAI 2021 · similarity 0.87)
- HiCI: Hierarchical Construction–Integration for Long-Context Attention (ICML 2026 · similarity 0.87)
- A Question-Focused Multi-Factor Attention Network for Question Answering (AAAI 2018 · similarity 0.86)
- The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentation (NeurIPS 2025 · similarity 0.86)
- Autofocus Retrieval: An Effective Pipeline for Multi-Hop Question Answering With Semi-Structured Knowledge (NeurIPS 2026 · similarity 0.85)
- SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking (NeurIPS 2026 · similarity 0.84)
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.