Research map / Large language models / Retrieval and knowledge
Translucent Answer Predictions in Multi-Hop 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 Global Layer over passages (LoGIX vs passage-level BERT) (ablation) | LoGIX supporting fact predictor · HotpotQA dev | Sup. Facts EM: 44.07 -> 53.11 | Helped |
| encode passages jointly rather than sentences alone (ablation) | LoGIX · HotpotQA dev | Sup. Facts EM: 27.70 -> 44.07 | Helped |
| add span selection pre-training (SSPT) (ablation) | Answer Predictor · HotpotQA dev | Answer EM: 61.62 -> 65.87 | Helped |
| add span selection pre-training (SSPT) (ablation) | LoGIX · HotpotQA dev | Sup. Facts F1: 85.27 -> 86.17 | Helped |
| use BERT-large instead of BERT-base (ablation) | Answer Predictor · HotpotQA dev | Answer EM: 63.31 -> 65.87 | Helped |
| use BERT-large instead of BERT-base (ablation) | LoGIX · HotpotQA dev | Sup. Facts F1: 85.13 -> 86.17 | Helped |
| train Answer Predictor on predicted supporting facts (joint training) (ablation) | Answer Predictor · HotpotQA dev | Answer EM: 65.16 -> 65.87 | Helped |
| raise supporting fact threshold from 0.1 to 0.4 (ablation) | TAP Answer Predictor · HotpotQA dev | Answer EM: 65.01 -> 64.32 | Hurt |
| ensemble multiple TAP models | TAP 2 · HotpotQA test | Answer EM: 64.99 -> 66.64 | Helped |
Closest papers by meaning
- Select, Answer and Explain: Interpretable Multi-Hop Reading Comprehension over Multiple Documents (AAAI 2020 · similarity 0.88)
- Answer, Assemble, Ace: Understanding How LMs Answer Multiple Choice Questions (ICLR 2025 · similarity 0.86)
- Question Calibration and Multi-Hop Modeling for Temporal Question Answering (AAAI 2024 · similarity 0.85)
- NovelQA: Benchmarking Question Answering on Documents Exceeding 200K Tokens (ICLR 2025 · similarity 0.85)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.85)
- Tree-of-Reasoning Question Decomposition for Complex Question Answering with Large Language Models (AAAI 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.