Latency-Aware Bid Acceptance under Operational Feasibility: A Public Benchmark with Hindsight Ceilings

The paper introduces FreightBidBench, a dependency-free and publicly calibrated benchmark for online truckload bid acceptance as a closed-loop stochastic decision problem. It makes pickup reachability, appointment windows, simplified hours-of-service constraints, stochastic yard delays, service penalties, fleet terminal value, and price-premium windows explicit and versioned. Using public Freight Analysis Framework and USDA truck-rate data, it defines LP and tighter per-truck Lagrangian information relaxations. Across ten seeds, a surrogate-rollout cascade recovers roughly 98% of rollout profit in tight- and scarce-capacity scenarios at 40–56% of rollout’s mean decision latency.