Data pruning and neural scaling laws: fundamental limitations of score-based algorithms

Open in new window