Revenue Maximization in Interdependent Value Settings
summary
Shows that a simple modification—reject each bidder with constant probability (here 1/2) then run Vickrey with reserves computed from survivors’ reported signals—gives a constant-factor approximation to optimal exposed‑IC revenue in interdependent settings with single‑crossing/additive values (proof uses the look‑ahead reserve argument and a two‑agent equal‑revenue counterexample to motivate rejection). Claims ~4.5× bound, extends to matroid constraints, and requires no distributional assumptions.
AI-written summary · about