A multimodal generative modeling approach combined with permutation-invariant set attention is investigated in this article to support long-wave infrared (LWIR) in-scene atmospheric compensation. The generative model can produce realistic atmospheric state vectors (T. H2O. O3) and their corresponding transmittance. https://cosmeticssquadets.shop/product-category/cleansing-soaps/