arXiv 2025 · Bi-Level Decision-Focused Causal Learning
Bi-DFCL：Bi-Level Decision-Focused Causal Learning for Large-Scale Marketing Optimization
用 RCT 的无偏 decision loss 校准 OBS 大样本训练，让 Bridge 负责纠偏，Target 负责最终上线。
论文精读索引
基于 Astro + MDX 的可复用论文精读阅读层。MDX 是源，HTML 是产物。当前样例为 DFCL / Bi-DFCL，框架本身与领域无关，可继续接入更多论文。
KDD 2024 · Decision-Focused Causal Learning
DFCL：Decision Focused Causal Learning for Direct Counterfactual Marketing Optimization
从 prediction-decision misalignment 出发，把 RCT 反事实估计、MCKP 预算优化、Lagrangian dual 与可训练 surrogate loss 合成一个营销决策训练闭环。