Clustered shading
Dividing the view frustum into a three-dimensional grid of cells — along depth too — and giving each cell its own light list, instead of Forward+'s flat screen tiles.
Forward+ divides only the screen in 2D, so lights at different distances inside one tile land in the same list and get shaded needlessly. Clustered shading adds division along depth: the frustum becomes a stack of cells (froxels), and a light is written only into the cells it actually intersects. The result is tighter light lists and, unlike Forward+, no dependence on a separate depth pre-pass.
Described by Olsson, Billeter and Assarsson (2012) in both a deferred and a forward variant. Doom (2016) and modern engines such as Unreal use it. The method works just as well with deferred shading as with forward, which was the whole point of the original paper.