An identification of soil parameters is performed by inverse analysis of two undrained triaxial cyclic tests on normally consolidated clay (kaolinite). A two-surface plasticity model is used for simulations and genetic algorithms are selected for the optimization procedure. First the problem of inverse analysis is formulated as a mono-objective problem, so that each test is considered separately and both sets of solutions are analyzed. Secondly, in order to take into account simultaneously the results of both tests, a multi-objective problem is considered and solved with a multi-objective genetic algorithm, which provides a set of equivalent solutions in terms of Pareto. Considering this set, a tradeoff is determined in order to fit at best the experimental and numerical curves for both tests and the relevance of the selected constitutive model is discussed.