PopQA
An open-domain question-answering benchmark dataset used to evaluate models' retrieval and generation capabilities on knowledge-intensive question-answering tasks.
Timeline (last 90 days)
At 64× compression, AdaMem's average relative gain expands to 14.6%, with a maximum improvement of 9.8 percentage points (19.7%) on the PopQA dataset.
PopQA is a question answering dataset constructed from long-tail entities in Wikipedia; a large proportion of its questions involve obscure knowledge, making it more dependent on retrieval quality.
All Facts (2)
At 64× compression, AdaMem's average relative gain expands to 14.6%, with a maximum improvement of 9.8 percentage points (19.7%) on the PopQA dataset.
50%UnverifiedPopQA is a question answering dataset constructed from long-tail entities in Wikipedia; a large proportion of its questions involve obscure knowledge, making it more dependent on retrieval quality.
50%