paper-with-me

홈 › Papers

Variance Reduction for Expectations with Diffusion Teachers

2026-05-20 · Jesse Bettencourt, Xindi Wu, Matan Atzmon, James Lucas, Jonathan Lorraine arxiv

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 introduce CARV, a compute-aware variance-accounting framework that motivates a hierarchical MC estimator: amortize the expensive upstream computation over cheap diffusion-noise resamples, sharpened by timestep importance sampling and a stratified-inverse-CDF construction. In our text-to-3D distillation and attribution experiments, CARV delivers 2-3x effective compute multipliers (most from amortized reuse; ~25% additional from IS+stratification) without changing the objective; in single-step distillation, the same techniques cut gradient variance by an order of magnitude but do not improve downstream FID, marking the regime where MC variance is no longer the bottleneck.

📄 PDF Abstract BibTeX arXiv:2605.21489

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

ARSM: Augment-REINFORCE-Swap-Merge Estimator for Gradient Backpropagation Through Categorical Variables

2019-05-04 · Mingzhang Yin, Yuguang Yue, Mingyuan Zhou

To address the challenge of backpropagating the gradient through categorical variables, we propose the augment-REINFORCE-swap-merge (ARSM) gradient estimator that is unbiased and has low variance. ARSM first uses variabl…

Estimating the Optimal Covariance with Imperfect Mean in Diffusion Probabilistic Models

2022-06-15 · Fan Bao, Chongxuan Li, Jiacheng Sun, Jun Zhu 외

Diffusion probabilistic models (DPMs) are a class of powerful deep generative models (DGMs). Despite their success, the iterative generation process over the full timesteps is much less efficient than other DGMs such as …

Computational Efficiency

SGD with Variance Reduction beyond Empirical Risk Minimization

2015-10-16 · Massil Achab, Agathe Guilloux, Stéphane Gaïffas, Emmanuel Bacry

We introduce a doubly stochastic proximal gradient algorithm for optimizing a finite average of smooth convex functions, whose gradients depend on numerically expensive expectations. Our main motivation is the accelerati…

Survival Analysis

AR-MAP: Are Autoregressive Large Language Models Implicit Teachers for Diffusion Large Language Models?

2026-02-02 · Liang Lin, Feng Xiong, Zengbin Wang, Kun Wang 외 arxiv

Diffusion Large Language Models (DLLMs) have emerged as a powerful alternative to autoregressive models, enabling parallel token generation across multiple positions. However, preference alignment of DLLMs remains challe…

Transfer Learning

Amortized variance reduction for doubly stochastic objectives

2020-03-09 · Ayman Boustati, Sattar Vakili, James Hensman, ST John

Approximate inference in complex probabilistic models such as deep Gaussian processes requires the optimisation of doubly stochastic objective functions. These objectives incorporate randomness both from mini-batch subsa…

Gaussian Processes