Support-Aware Offline Policy Selection for Ad Auctions
A new framework for offline reserve-price evaluation in advertising auctions addresses the risk of misleading replay tables. The method, described in arXiv:2605.21736, converts logged evidence into a conservative decision object that distinguishes certified policies from statistically dominated alternatives and unresolved candidates. The main theoretical result provides a unified finite-catalog guarantee under simultaneous uncertainty, offering a more robust approach to policy selection without requiring a single point-estimate winner.
Key facts
- arXiv:2605.21736 introduces a support-aware offline decision framework for reserve-policy selection in advertising marketplaces.
- Logged advertising auctions make offline reserve-price evaluation attractive but risky.
- Replay tables can identify policies with large apparent yield gains but may hide weak threshold support, multiple-comparison effects, subgroup harm, and bidder-response uncertainty.
- Existing replay and off-policy evaluation methods estimate or rank policy values but do not directly answer whether evidence justifies validation.
- The framework outputs a conservative decision object consisting of certified policies, statistically dominated alternatives, and unresolved candidates.
- The main theoretical result gives a unified finite-catalog guarantee under simultaneous uncertainty.
Entities
—