The paper tackles a classical question in algebraic geometry: how much geometric information about a space can be recovered from another space built from it? The specific setting involves moduli spaces, which are geometric objects that parametrize all vector bundles of a given rank and degree over an algebraic curve. A vector bundle is a structured family of vector spaces attached to points of a curve in a compatible way, and collecting all such bundles into a single geometric object gives the moduli space. The central question is a Torelli-type problem: if the product of several such moduli spaces is isomorphic to the product of another collection, can you conclude that the original pieces were essentially the same?
The main result confirms that the answer is yes, at least for curves of genus four or higher. The authors prove that any isomorphism between products of these moduli spaces forces a matching between the individual factors, up to reordering and obvious symmetries. This is a strong rigidity statement. The key technical tool is a decomposition theorem the authors develop for products of a broad class of geometric spaces, namely normal projective varieties with Picard rank one and discrete Picard group. These conditions roughly mean the spaces have very few line bundles, which limits how isomorphisms can mix the factors.
The result is proved both for the moduli spaces themselves and for the related moduli stacks, which are more refined objects that keep track of the symmetries of each bundle rather than treating isomorphic bundles as identical points. As a concrete payoff, the authors compute the full automorphism group of these products, meaning they classify all self-isomorphisms of the product. This completes a picture that was previously understood only for individual moduli spaces, extending it to products and giving a precise structural answer about their symmetries.