[Objective] Rock-filled concrete (RFC) is inevitably subjected to repeated loading-unloading processes during service. However, studies on the stiffness degradation and hysteretic behavior of RFC under cyclic loading remain limited. This study investigated the mechanical response of RFC under cyclic loading through systematic uniaxial cyclic compression tests, quantified the evolution of stiffness degradation and hysteretic energy, compared these characteristics with those of normal concrete, and developed a hysteretic constitutive model for RFC. [Methods] Cubic specimens with a side length of 200 mm were used. The strength grades of self-compacting concrete (SCC) (25 MPa and 40 MPa) and rockfill ratios (55% and 0%, i.e., plain SCC) were selected as the main variables. Uniaxial cyclic compression tests were conducted. To address brittle post-peak failure caused by insufficient stiffness of the testing machine, a synergistic loading system incorporating an auxiliary beam and jacks was introduced, enabling multi-level controlled unloading and reloading after peak stress. Externally mounted displacement sensors and a load cell were used to record complete stress-strain curves in real time, and the equivalent elastic modulus and hysteretic energy for each cycle were calculated. The test results were then normalized and compared with classical data for normal concrete reported in the literature. To overcome the limitations of the classical Sima hysteretic model, an improved constitutive model incorporating a nonlinear correction for the unloading branch and a bilinear representation for the reloading branch was proposed. [Results] (1) The elastic modulus of RFC exhibited a gradual degradation trend with an increasing number of cycles. Specimens with higher SCC strength showed a higher initial modulus but a faster degradation rate. The hysteretic energy at each loading level also decreased gradually, and the specimens with higher SCC strength exhibited greater energy dissipation capacity. (2) Compared with normal concrete, RFC exhibited a lower initial modulus but a significantly slower stiffness degradation rate. The post-peak descending segment of the normalized skeleton curve was steeper, indicating lower ductility. The hysteretic loops of RFC were flatter and crescent-shaped, with significantly lower fullness than those of normal concrete. The normalized loop area at the same loading level was only 55%-65% of that of normal concrete, and the accumulation rate of plastic residual strain was also slower, suggesting that the rockfill skeleton exerted a confining effect on the matrix plastic flow. (3) The classical Sima hysteretic model did not adequately fit the behavior of RFC because it exhibited excessive curvature in the low-stress region of the unloading branch and failed to represent the stiffness transition during reloading. The loop-area prediction error reached 18.5%. The improved model, incorporating a stress correction at the end of unloading and a bilinear reloading description, reduced the average fitting errors of the unloading and reloading branches from 12.7% to 4.3% and the loop-area error to 6.2%, thereby improving the accuracy in characterizing the hysteretic behavior of RFC. [Conclusion] RFC under uniaxial cyclic compression exhibits stiffness degradation and hysteretic behavior distinct from those of normal concrete, which is primarily attributed to the dominant elastic load-bearing role of the rockfill skeleton and its restraining effect on matrix damage propagation. Although increasing SCC strength can enhance initial stiffness and energy dissipation capacity, it also accelerates stiffness degradation, requiring careful consideration in design. The proposed improved hysteretic constitutive model accurately describes the unloading nonlinearity and reloading stiffness evolution of RFC, addresses the curvature distortion in the low-stress region and reloading stiffness discontinuity in classical models, and provides an effective material model for nonlinear dynamic analysis of RFC structures. This study deepens the understanding of cyclic damage evolution in RFC composite materials at the mesoscopic level by revealing the differences between RFC and normal concrete in terms of envelope characteristics, hysteresis loop fullness, and plastic accumulation behavior. It can provide experimental and theoretical bases for improving relevant specifications and safety assessment in engineering applications.