A novel method named XTNet has been developed to address the challenges inherent in counterfactual causal inference involving multi-category, multi-valued treatments. Traditional models in this domain have predominantly concentrated on binary treatments, which limits their applicability to more complex intervention scenarios. XTNet overcomes these limitations by providing a scalable and effective framework capable of handling the intricacies of multiple treatment categories and values. This advancement is particularly significant as it broadens the scope of causal inference methods beyond the binary treatment paradigm. By leveraging dynamic neural masking, XTNet enhances the estimation of cross-treatment effects, facilitating more accurate evaluations of diverse interventions. This approach represents a meaningful step forward in causal inference research, as highlighted in recent discussions on the topic. The introduction of XTNet thus marks a promising development for applications requiring nuanced analysis of multi-valued treatment effects.