Research map / Large language models / Retrieval and knowledge
QASC: A Dataset for Question Answering via Sentence Composition
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 two-step retrieval instead of single-step retrieval | BERT-MCQ on QASC · QASC dev, original 4-way dataset | accuracy: 78.2 -> 86.4 | Helped |
| Use two-step IR instead of single-step IR | Fact retrieval on QASC · QASC | gold-fact pair recall at top 10: 2.9% -> 44.4% | Helped |
| Add adversarial distractors instead of random distractors | BERT-MCQ with two-step retrieval · QASC dev | accuracy: 83.3 -> 72.9 | Hurt |
| Add random distractors to make the questions 8-way (ablation) | BERT-MCQ with two-step retrieval · QASC dev | accuracy: 86.4 -> 83.3 | No clear effect |
| Use the larger ARC corpus with two-step retrieval (ablation) | BERT-MCQ on QASC · QASC test | accuracy: 67.0 -> 58.3 | Hurt |
| Use the larger ARC corpus with single-step retrieval (ablation) | BERT-MCQ on QASC · QASC test | accuracy: 53.2 -> 57.0 | Helped |
| Replace BERT-LC with whole-word-masking BERT (ablation) | BERT-MCQ with two-step retrieval and RACE + SCI fine-tuning · QASC test | accuracy: 68.5 -> 73.2 | Helped |
| Increase the number of adversary models (ablation) | Multi-adversary distractor selection · QASC | fine-tuned model score | No clear effect |
| Add RACE + SCI fine-tuning (ablation) | BERT-MCQ with two-step retrieval · QASC test | accuracy: 67.0 -> 68.5 | Helped |
Closest papers by meaning
- Tree-of-Reasoning Question Decomposition for Complex Question Answering with Large Language Models (AAAI 2024 · similarity 0.90)
- MEQA: A Benchmark for Multi-hop Event-centric Question Answering with Explanations (NeurIPS 2024 · similarity 0.89)
- Question Calibration and Multi-Hop Modeling for Temporal Question Answering (AAAI 2024 · similarity 0.88)
- CofCA: A STEP-WISE Counterfactual Multi-hop QA benchmark (ICLR 2025 · similarity 0.87)
- Cooperative Retrieval-Augmented Generation for Question Answering: Mutual Information Exchange and Ranking by Contrasting Layers (NeurIPS 2025 · similarity 0.87)
- Autofocus Retrieval: An Effective Pipeline for Multi-Hop Question Answering With Semi-Structured Knowledge (NeurIPS 2026 · similarity 0.87)
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.