AI RESEARCH
Additive Control Variates Dominate Self-Normalisation in Off-Policy Evaluation
arXiv CS.LG
•
ArXi:2602.14914v2 Announce Type: replace Off-policy evaluation (OPE) is essential for assessing ranking and recommendation systems without costly online interventions. Self-Normalised Inverse Propensity Scoring (SNIPS) is a standard tool for variance reduction in OPE, leveraging a multiplicative control variate. Recent advances in off-policy learning suggest that additive control variates (baseline corrections) may offer superior performance, yet theoretical guarantees for evaluation are lacking.