Highly coarse-grained models for investigating the self-assembly of lipids and copolymer materials are discussed. Soft interactions between segments that represent many atoms naturally arise in the course of systematic coarse-graining, and they are necessary for modeling fluctuation effects whose strengths is dictated by a large invariant degree of polymerization. The soft non-bonded interactions of the coarse-grained models are related to the excess free-energy functional of an equivalent field-theoretic description. The connection between the particle-based model and the field-theoretic description helps to identify the physical significance of the model interactions. Non-bonded interactions, which describe the complex phase behavior of compressible mixtures or include local fluid-like packing effects of the coarse-grained segments, can be systematically constructed based on liquid-state theory or classical density functional theory. Details of the computational implementation and limitations of soft coarse-grained models are discussed. Two computational techniques-field-theoretic force-matching and umbrella sampling-are devised for computing a free-energy functional from a particle-based description. They can be employed to (i) derive the non-bonded free-energy functional of a soft coarse-grained model from a more detailed computational model or to (ii) derive a field-theoretic description from a particle-based model. Moreover, different strategies for accurately calculating free energies of self-assembled systems are described and selected applications presented.