DACRI: Decision-Aware Causal Intervention Ranking for Critical Supply Chains
The paper argues that supply-chain response should be ranked by recoverable net value, not just disruption detection.
The authors introduce CriticalSCM-Bench v1, a synthetic benchmark with causal ground truth, counterfactual rollouts, and an explicit net-value target. LambdaMART improves median normalized net value against a static full-information benchmark in semiconductor and critical-material cases, but not in digital infrastructure. There, a constant-buffer policy performs better, undercutting any blanket case for more complex models. Stress tests show policy rankings can shift with intervention fidelity, timing, cost, and unseen disruptions. ArXiv · AI/CL/LG's note
The authors introduce CriticalSCM-Bench v1, a synthetic benchmark with causal ground truth, counterfactual rollouts, and an explicit net-value target. LambdaMART improves median normalized net value against a static full-information benchmark in semiconductor and critical-material cases, but not in digital infrastructure. There, a constant-buffer policy performs better, undercutting any blanket case for more complex models. Stress tests show policy rankings can shift with intervention fidelity, timing, cost, and unseen disruptions. ArXiv · AI/CL/LG's note
score 4