List your company
ResearchFormerly WizWand

Off-policy model selection on KuaiRec

28Top-1 AccuracyTree-DR
0.967.981522.02May 28, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Tree-DRPropensity source=Forward-DP-enabled, K=4, Nlog=10002026.05280.420.056
Tree-WISPropensity source=Trajectory IS, K=4, Nlog=10002026.05260.240.059
DP-MPL-WISPropensity source=Forward-DP-enabled, K=4, Nlog=10002026.05260.220.075
FF-WISPropensity source=Forward-flow IS, K=4, Nlog=10002026.05220.270.058
FF-DRPropensity source=Forward-flow IS, K=4, Nlog=10002026.05180.350.072
DP-MPL-OISPropensity source=Forward-DP-enabled, K=4, Nlog=10002026.0510-0.530.369
FF-OISPropensity source=Forward-flow IS, K=4, Nlog=10002026.058-0.450.204
DP-OPCB-DRPropensity source=Forward-DP-enabled, K=4, Nlog=10002026.0580.380.051
Tree-OISPropensity source=Trajectory IS, K=4, Nlog=10002026.056-0.530.209
DP-OPCB-OISPropensity source=Forward-DP-enabled, K=4, Nlog=10002026.0520.090.09