Abstract: Conglomerate reservoirs exhibit complex lithology, extreme heterogeneity, high gravel content, cross-scale particle-size distributions, and spatially irregular gravel occurrence. In hydraulic fracturing, these characteristics strongly interfere with hydraulic fracture propagation path, resulting in complex fracture geometries and low predictive accuracy of fracture parameters, thereby limiting the theoretical support for fracturing design in such reservoirs. To address this challenge, a hydro-mechanically coupled numerical model was established for fracturing in conglomerate reservoirs based on the extended finite element method, in which adaptive mesh refinement method was introduced to explicitly capture centimeter-sized gravel clasts. The model enables accurate cross-scale simulation and characterization of fracture propagation from centimeter to hectometer size and was validated using available fracture propagation data for conglomerate formations. On this basis, meter-sized simulations were performed to clarify the mechanisms and controlling factors of fracture–gravel interaction, and these results were subsequently extended to reservoir-scale simulations to investigate fracture propagation patterns in conglomerate reservoirs. The results indicate that fracture–gravel interaction can be categorized into four main modes: bypassing, arresting, penetrating, and induced branching. Injection rate, fluid viscosity, horizontal stress differential, and gravel content are identified as the dominant factors governing interaction mode and fracture propagation geometies. Reservoir-scale simulations further reveal that natural fractures primarily determine the propagation path and geometies of hydraulic fractures, whereas gravel clasts exert only a minor influence on the overall fracture pattern. Key geological factors controlling fracture network development, in descending order of significance, are natural fracture density, natural fracture length, horizontal stress differential, injection rate, fluid viscosity, and gravel density. This study provides a theoretical foundation for the optimization of hydraulic fracturing parameters in conglomerate reservoirs and offers valuable engineering guidance for enhancing stimulation performance and development efficiency in these complex formations.