We study variance reduction for score estimation and diffusion based sampling in settings where the clean (target) score is available or can be approximated. We start from the Target Score Identity (TSI), which expresses the noisy marginal score as a conditional expectation under

the forward diffusion kernel. Building on this, we develop: (i) a nonparametric estimator based on self normalized importance sampling that can be used directly with standard solvers (ii) a varianceminimizing state and time dependent blending rule between Tweedie type and TSI estimators together with an anticorrelation analysis, (iii) a data-only extension based on locally fitted proxy scores, and (iv) a likelihood informed extension to Bayesian inverse problems. Experiments on synthetic targets and PDE governed inverse problems demonstrate improved sample quality for a fixed simulation budget.