The paper studies a special class of complex algebraic surfaces, called surfaces of general type with certain numerical invariants fixed (specifically, the self-intersection of the canonical class equals 1 and the geometric genus equals 1). These are among the simplest and most mysterious surfaces in algebraic geometry. A classical result by Catanese gives a concrete way to write down equations for these surfaces inside a weighted projective space, and it is already known that the "global Torelli theorem" fails for them, meaning you cannot always recover the surface just from certain topological/algebraic data called the Hodge structure.
The authors first tackle the question of how many symmetries such a surface can have. They describe the full automorphism group in terms of the coefficients appearing in Catanese's normal form for the equations, and prove that the number of automorphisms is at most 72. They then identify exactly which surfaces achieve this maximum: two surfaces with mild singularities and a whole one-parameter family of smooth surfaces called X_r. For this family, they go further and explicitly compute a basis for a key piece of the surface's cohomology (called the primitive cohomology) using residues of explicit polynomials, and they work out the intersection form in that basis. This is a concrete and computationally involved piece of work that makes the abstract Hodge theory very explicit.
Finally, the authors investigate how the symmetries of the surface relate to symmetries of its Hodge structure. There is a natural map sending each geometric automorphism of X_r to a symmetry of the associated lattice of cohomology classes (a Hodge isometry). The authors show this map is injective, so no geometric symmetry acts trivially on cohomology, but its image is an extraordinarily small fraction of all Hodge isometries: the index is at least 331,776. This means there are vastly many cohomological symmetries that do not come from any geometric symmetry of the surface. This is a strong and quantitative form of failure of what the authors call the "Torelli principle for automorphisms," showing a fundamental gap between geometry and Hodge theory for this family of surfaces.