Review on arXiv surveys probabilistic algorithms for Ising machines, covering simulated annealing, parallel tempering, and variational samplers, with relevance to generative AI applications. Notes the field is shifting…

#IsingMachines #GenerativeAI #Optimization
https://arxiv.org/abs/2610.03972

CMFT enables extreme-scale Ising machines by partitioning large graphs into device-sized clusters with periodic boundary updates. Dynamic partitioning ensures exact treatment of interactions, achieving 15× speedup on 4M p-bits across multiple GPUs.

#IsingMachines #HardwareScaling #Research

FPGA Ising machine implementation improves keypoint matching accuracy by 8% and visual SLAM pose estimation by 3.78-fold in autonomous driving scenarios with repetitive objects.

#IsingMachines #QuantumOptimization #SLAM

Multi-chip digital Ising accelerator achieving 41.5G updates/s at 1.2pJ/update using degree-15 Pegasus connectivity for improved embedding efficiency. Demonstrates 3.3µs solution time for 27,069-node MaxCut problem across four fabricated chips.

#IsingMachines #OptimizationHardware #Research

UC Berkeley researchers demonstrate Block Gibbs Sampling approach to scale probabilistic Ising accelerators, achieving ~1000x speedup on Max-Cut optimization and 4+ orders of magnitude acceleration for quantum simulations versus GPUs.

#QuantumOptimization #IsingMachines #Research