AI RESEARCH
Variance Reduction for Expectations with Diffusion Teachers
arXiv CS.LG
•
ArXi:2605.21489v1 Announce Type: new Pretrained diffusion models serve as frozen teachers feeding downstream pipelines such as text-to-3D, single-step distillation, and data attribution. The teacher gradients these pipelines consume are Monte Carlo (MC) expectations over noise levels and Gaussian noise samples; their estimator variance dominates compute cost because each draw requires expensive upstream work (rendering, simulation, encoding). We