common approach to RNA folding pipelines is to start by predicting secondarystructures from sequence, and then, tertiary spatial folds from the secondary structure. Inthis review, we are interested in a backward approach: we explore what information fromknown solved RNA 3D structures can be used to improve secondary structure prediction.We propose a Pareto-based method for predicting secondary structures by minimizing abi-objective half energy-based, half knowledge-based potential. The tool outputs the sec-ondary structures from the Pareto set. We use it to compare several approaches to insertRNA modules into the secondary structures and benchmark them against the RNAstrandsecondary structure database. We compare two dierent module data sources, Rna3Dmotifand The RNA Motif Atlas and dierent ways to score the module insertions taking intoaccount module size, module complexity, or module probability according to models likeJAR3D and the recent BayesPairing method.