Computational thinking (CT) is a critical 21st-century skill that equips undergraduate students to solve problems systematically and think algorithmically. A key component of CT is computational creativity, which enables students to generate novel solutions within programming constraints. Humanoid robots are increasingly explored as promising tools to enhance CT skills, fostering teamwork and creativity in collaborative settings. However, gaps remain in understanding how different learning modalities impact the development of these skills. This study examines the comparative effects of robot-guided and simulation-based collaborative learning on undergraduate students' computational creativity and CT skills. The study involved 71 undergraduate students, divided into small groups and randomly assigned to begin with either a robot-guided or simulation-based modality, switching to the alternate modality in the following session. Data were collected through group log data and pre- and post-intervention questionnaires. The results indicated that the robot-guided modality significantly enhanced computational creativity in terms of originality and elaboration, while both modalities supported flexibility equally. Additionally, students reported higher CT skills following the robot-guided activity, with the most notable improvements in cooperation and creativity. Lastly, fewer group interaction difficulties were reported during the robot-guided activity, supporting its value for collaborative learning. These findings highlight humanoid robots as a valuable complement to virtual learning environments, offering unique opportunities to foster creativity, collaboration, and problem-solving in undergraduate education.