HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
About
Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. We introduce HotpotQA, a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) we provide sentence-level supporting facts required for reasoning, allowing QA systems to reason with strong supervision and explain the predictions; (4) we offer a new type of factoid comparison questions to test QA systems' ability to extract relevant facts and perform necessary comparison. We show that HotpotQA is challenging for the latest QA systems, and the supporting facts enable models to improve performance and make explainable predictions.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multi-hop Question Answering | HotpotQA fullwiki setting (test) | Answer F134.4 | 64 | |
| Question Answering | HotpotQA distractor (dev) | Answer F158.3 | 45 | |
| Multi-hop Question Answering | HotpotQA fullwiki setting (dev) | Answer F134.36 | 38 | |
| Question Answering | HotpotQA (test) | Ans F132.9 | 37 | |
| Question Answering | HotpotQA distractor setting (test) | Answer F159.02 | 34 | |
| Question Answering | HotpotQA full wiki (dev) | F134.4 | 20 | |
| Supporting Fact Prediction | HotpotQA full wiki (dev) | F1 Score41 | 19 | |
| Supporting Fact Prediction | HotpotQA distractor (dev) | F1 Score66.7 | 13 | |
| Question Answering | HotpotQA Full Wiki hidden (test) | F132.9 | 12 | |
| Supporting Facts Prediction | HotpotQA Full Wiki hidden (test) | F1 Score37.7 | 11 |