phenomenon · first seen 18 Oct 2026
benchmark data leakage
Also: train-test contamination
Overlap between training and test splits of a benchmark that inflates reported evaluation scores.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Only Means Only: what a harmonization benchmark score averages overPapers · 18 Oct 2026 · Imad Buljić, Delaida Muminović
Latest sources
- Only Means Only: what a harmonization benchmark score averages overPapers · 18 Oct 2026 · Imad Buljić, Delaida Muminović