Highly sensitive RNA detection at ultralow concentrations remains challenging for conventional surface plasmon resonance (SPR) biosensors because low-molecularI biomolecules induce only weak effective refractive-index perturbations within the limited evanescent-field penetration depth. Here, we report a dual-scale gold nanoparticle (AuNP)−enhanced fiber-optic SPR biosensor that overcomes this intrinsic limitation through a mechanism-guided sandwich-type architecture. The sensor employs a multimode fiber−single-mode fiber−multimode fiber (MMF−SMF−MMF) configuration to enhance evanescent-field excitation at the metal−dielectric interface. Plasmonic amplification is achieved by decoupling probe immobilization and signal enhancement into two AuNP populations with distinct roles: large AuNPs functionalized with probe DNA form a stable, high-density recognition layer on the gold-coated fiber surface, while small AuNPs conjugated with target RNA act as mobile amplification units. Upon specific DNA−RNA hybridization, a hierarchical plasmonic sandwich structure is formed within the effective sensing volume, enabling synergistic enhancement through near-field electromagnetic coupling and mass-loading effects. Finite-element simulations and experiments consistently demonstrate substantially stronger localized electromagnetic fields and higher refractive-index sensitivity compared with bare gold films and monolayer AuNP-modified SPR sensors. Importantly, the assay does not rely on nucleic-acid amplification reactions or fluorescent labeling. The sensor exhibits a linear response to RNA concentrations from 0 to 700 fM, achieving a sensitivity of 6.5 pm/fM and a detection limit of 97 fM, corresponding to a 7.5-fold signal enhancement over conventional single-layer AuNP-based SPR sensors. This work demonstrates that a dual-scale AuNP assisted sandwich architecture can effectively amplify weak SPR responses from thiol-modified model RNA targets without requiring nucleic-acid amplification reactions. Further optimization toward direct detection of unmodified RNA in real biological samples will be pursued in future work.