Preprint
Aug 2026
Excess Separability: Nuisance-Controlled Residual-Stream Probing for Benchmark Contamination Detection
It is shown the natural way to do this does not work, specify one that survives measurement, then finds that the correction making it work carries more variance than the null it is tested against, and that the correction making it work carries more variance than the null it is tested against.
Floriane C. M. Braun
· 0 citations