Inject Where It Matters: Objective-Dependent Feature Control in Rectified-Flow Image Editing

Many prompt-driven FLUX editing methods rely on injecting source features into target-conditioned generation, creating an inherent trade-off between source preservation and the desired edit. We address this problem by controlling where source features are injected and which transformer blocks are used. First, we introduce adaptive masks that spatially gate feature injection using differences between source- and target-conditioned block features at the same states along inversion. Second, we use REINFORCE to select Top-$K$ injection blocks under different preservation--editability objectives. The selected blocks are fixed at inference and require no online search. On PIE-Bench with FLUX.1-dev, spatial masks provide direct control over the preservation--editability trade-off, while nine learned blocks can retain the preservation achieved with eighteen. The learned policies reveal objective-dependent block preferences: preservation-focused optimization concentrates on a compact group of blocks, while increasing the weight on target-prompt alignment changes which blocks are selected.