Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces diffusion-guided estimators for counterfactual distribution learning in high-dimensional settings, addressing the limitations of isotropic smoothing by adapting to the underlying lower-dimensional structure. They propose diffusion-informed smoothing for counterfactual densities and diffusion-informed score smoothing for counterfactual scores, combining causal nuisance adjustment with geometry-adaptive localization. Theoretical results demonstrate that the leading stochastic error is governed by an effective dimension induced by the diffusion-guided kernel, and experiments on CelebA confirm improved error decay compared to standard methods.
By aligning smoothing with local outcome geometry, diffusion-guided estimators unlock more stable and efficient counterfactual inference in high-dimensional spaces.
We study counterfactual distribution learning for high-dimensional outcomes whose counterfactual law may concentrate near lower-dimensional structure. Standard isotropic smoothing treats all ambient directions equally, leading to unfavorable scaling and unstable local inference. We propose two diffusion-guided estimators based on semiparametric debiasing: diffusion-informed smoothing for counterfactual densities and diffusion-informed score smoothing for counterfactual scores. The estimators combine causal nuisance adjustment with geometry-adaptive localization driven by diffusion score information, removing first-order nuisance bias while aligning smoothing with local outcome geometry. We establish asymptotic expansions, risk bounds, and inference procedures for smoothed density and score-based targets, with ambient density inference obtained under additional approximation conditions. Under structural geometry conditions, the leading stochastic error is governed by an effective dimension induced by the diffusion-guided kernel, rather than by the ambient dimension. Semi-synthetic experiments based on CelebA show steeper error decay for geometry-adaptive methods, supporting the proposed effective-dimension theory.