Correlated Diffusion replaces independent noise with structured MCMC dynamics, enabling generative modeling on hyper-efficient probabilistic computers.
March 31, 2026
Original Paper
From Independent to Correlated Diffusion: Generalized Generative Modeling with Probabilistic Computers
arXiv · 2603.27996
The Takeaway
By moving the computational burden from the neural network to a structured stochastic kernel (Ising couplings), this framework achieves orders-of-magnitude improvements in energy efficiency and throughput. It represents a fundamental shift toward hardware-aware generative modeling using p-bits instead of GPUs.
From the abstract
Diffusion models have emerged as a powerful framework for generative tasks in deep learning. They decompose generative modeling into two computational primitives: deterministic neural-network evaluation and stochastic sampling. Current implementations usually place most computation in the neural network, but diffusion as a framework allows a broader range of choices for the stochastic transition kernel. Here, we generalize the stochastic sampling component by replacing independent noise injectio