When Does Data Value Reduce to Class Balance? A Coverage View of Per-Point Data Valuation
arXiv:2610.34801v1 Announce Type: cross Abstract: Per-point data-valuation scores -- Data Shapley, influence, leverage -- are routinely used to select training data: keep the points with the top $k$ scores. We argue that top-$k$ by a fixed, mode-blind per-point score is the wrong selection rule pre…
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-10-28 04:00 · arXiv stat.ML
When Does Data Value Reduce to Class Balance? A Coverage View of Per-Point Data Valuation