1 articles with this tag
A new theoretical framework reveals an inescapable trade-off between efficiency, compactness, and recall in long sequence models.
