L2L4 is an experimental 3x3 speedsolving method, the 4th proposal in a series of L2L methods proposed by Duncan Dicks. In this proposal, there is a system for completing the Last 2 Layers in 4 steps. These steps can be arranged however, but are usually referenced by how they affect the last layer. In a beginner layer-by-layer method, once completed the first two layers, one does CO, CP, EO, and EP, in a certain sequence, completing the final layer with these four steps (often combined when a cuber is more advanced/experienced). What 'pure' L2L4 aims to do is to simply build a layer, namely on D, then complete the aforementioned steps all the while completing the four medges (middle-edges).
L2L4 depends on efficient block-building for the First Layer, fast recognition for the four L2L algorithms, decent algorithm execution, and dedication to memorize the required algorithms.