In this talk, I will present an iPEPS study targeting the Neel Phase of the paradigmatic model of frustrated magnetism - J1-J2 model on the square lattice. The recent progress in both optimization and data analysis of iPEPS through finite correlation-length scaling allowed us to obtain accurate estimate of magnetization curve, vanishing around J2_c ~ 0.46(1) into a paramagnetic phase. First, I will briefly introduce the concepts behind gradient optimization of iPEPS with automatic differentiation and then proceed to the specifics pertinent to J1-J2 model. I will show how the correct U(1) symmetry classes can be inferred from the analysis of unrestricted simulations allowing for resolution of distinct families of local minima which otherwise complicate the data analysis. Finally, I will present the results of the AD optimizations of U(1)-symmetric iPEPS putting finite correlation-length scaling to use and discuss the future directions.