TILT: Model-Intrinsic Reward Alignment For Compositional Diffusion

Debottam Dutta, Jaehoon Hahm, Jianchong Chen +1

Consider conditional generation $p(x\mid C=\{c_1,c_2,\dots c_k\})$ where $C$ is a prompt composed of multiple concepts $c_i$. Diffusion models often struggle with compositional…