Implicit neural fields, usually encoded by a multilayer perceptron (MLP) that maps from coordinates (e.g., xyz) to indicators (e.g., signed distances), have proven outstanding promise as a high-fidelity and compact illustration. Nevertheless, the shortage of a daily and specific grid construction additionally makes it difficult to use generative modeling immediately on implicit neural fields as a way to synthesize new information. To this finish, we suggest HyperDiffusion, a novel strategy for unconditional generative modeling of implicit neural fields. HyperDiffusion operates immediately on MLP weights and generates new neural implicit fields encoded by synthesized MLP parameters. Particularly, a group of MLPs is first optimized to faithfully characterize particular person information samples. Subsequently, a diffusion course of is skilled on this MLP weight area to mannequin the underlying distribution of neural implicit fields. HyperDiffusion permits diffusion modeling over a implicit, compact, and but high-fidelity illustration of advanced indicators throughout 3D shapes and 4D mesh animations inside one single unified framework.