Researchers have developed a new deep learning framework called CIHSI-Net for estimating treatment effects from observational data. The framework uses a novel balancing method called Barycentric Fused Gromov-Wasserstein Balancing (BFG-WB) to reduce estimation variance and preserve local proximity structures. This is particularly useful in scenarios with multiple simultaneous treatments, such as marketing campaigns. Simulation studies show that CIHSI-Net outperforms existing methods, and an application to real-world data demonstrates its practical utility. --- Why it matters: This matters because it provides a more accurate way to analyze the effects of different treatments in complex scenarios, which is crucial for decision-making in fields like marketing and healthcare. Source: https://arxiv.org/abs/2608.22024