Abstract: An inversion method of facies configuration based on gradient structure tensor is proposed for small-scale karst reservoirs with complex fracture-vug development and interstitial materials in the Maokou Formation, the Sichuan Basin. The workflow consists of three key steps: facies-related structural configuration characterization using structure attributes derived from gradient structure tensor and its eigenvalues, initial model building with facies-related petrophysical properties derived from the geologic model and log data, and stochastic inversion constrained by the reservoir configuration model for reservoir prediction. The case study shows that gradient structure tensor can effectively describe weak chaotic reflections of small karst caves, enabling precise karst pinpointing and extraction. Configuration modeling based on gradient structure tensor can accurately characterize karst reservoirs of different patterns. The configuration constraint facilitates quantitative analysis and prediction of karst reservoirs.