NNiT: Width-Agnostic Neural Network Generation with Structurally Aligned Weight Spaces

Researchers introduced Neural Network Diffusion Transformers (NNiTs), a new approach that generates neural network parameters in a width-agnostic manner by treating weight matrices as tokenized patches. The method achieves over 85% success on unseen network architectures in robotics tasks, solving key challenges in generative modeling of neural networks.

• →NNiTs generate neural network weights without being tied to specific architecture dimensions by tokenizing weight matrices into patches.
• →Graph HyperNetworks with CNN decoders create structural alignment in weight spaces needed for patch-based processing.
• →The approach jointly models discrete architecture tokens and continuous weight patches in a single sequence model.
• →NNiT achieved >85% success rate on unseen architecture topologies in ManiSkill3 robotics tasks.
• →The method addresses permutation symmetries that complicate neural network parameter generation.