The production of light nuclei in relativistic heavy-ion collisions is well described by both the thermal model, where light nuclei are in equilibrium with hadrons of all species present in a fireball, and by the coalescence model, where light nuclei are formed due to final-state interactions after the fireball decays. We present and critically discuss the two models and further on we consider two proposals to falsify one of the models. The first proposal is to measure a yield of exotic nuclide 4Li and compare it to that of 4He. The ratio of yields of the nuclides is quite different in the thermal and coalescence models. The second proposal is to measure a hadron-deuteron correlation function which carries information whether a deuteron is emitted from a fireball together with all other hadrons, as assumed in the thermal model, or a deuteron is formed only after nucleons are emitted, as in the coalescence model. The p − 3He correlation function is of interest in context of both proposals: it is needed to obtain the yield of 4Li which decays into p and 3He, but the correlation function can also tell us about an origin of 3He.