Multi-step Reasoning
Complex reasoning requiring multiple inference steps (HotpotQA).
Benchmarks & Datasets
HotpotQA
113K question-answer pairs requiring reasoning over multiple Wikipedia documents.
N/A
Images
2018
Year
See Leaderboard
StrategyQA
2,780 yes/no questions requiring implicit multi-step reasoning to answer.
N/A
Images
2021
Year
See Leaderboard
GPQA
448 expert-level questions in biology, physics, and chemistry. Designed to be unsearchable.
N/A
Images
2024
Year
See Leaderboard