Beyond Overlap: Estimating the Causal Effect of Benchmark Exposure

Excerpt

arXiv:2609.27176v1 Announce Type: new Abstract: Evidence that evaluation material entered training does not reveal how much it affected evaluation. This distinction leaves a contaminated benchmark score difficult to interpret: provenance can establish contact, but only a counterfactual can quantify the performance attributable to that contact. We present LeakScale, an interventional framework for estimating this missing quantity. LeakScale creates fresh executable tasks that require private, fam