First-Order Efficiency for Probabilistic Value Estimation via A Statistical Viewpoint
ICML 2026 NExT-Game Workshop, 2026
This work studies probabilistic value estimation from a statistical viewpoint. It shows that several existing estimators share a common first-order error structure and uses that insight to design a more efficient surrogate-adjusted estimator.
Recommended citation: Liu, Z., Lee, K., Zhang, Y., and Tang, W. (2026). "First-Order Efficiency for Probabilistic Value Estimation via A Statistical Viewpoint." ICML 2026 NExT-Game Workshop.
Download Paper