NQ-Open
An open-domain question answering benchmark dataset released by Google, constructed from real user search queries. The questions are of high difficulty and are commonly used to evaluate the factual retrieval and question answering capabilities of language models.
Timeline (last 90 days)
On the NQ-Open dataset, the answer selector outperformed the filter on threshold average utility, but underperformed the "always abstain" baseline strategy.
NQ-Open is derived from real Google search queries; the questions are more natural and more ambiguous, and overall accuracy is generally lower than on TriviaQA.
All Facts (2)
On the NQ-Open dataset, the answer selector outperformed the filter on threshold average utility, but underperformed the "always abstain" baseline strategy.
50%UnverifiedNQ-Open is derived from real Google search queries; the questions are more natural and more ambiguous, and overall accuracy is generally lower than on TriviaQA.
50%