Despite the recognized link between scientific reasoning, mental sets, and conceptual change, the cognitive mechanisms underlying successful conceptual change remain unclear. To explore this, we developed computer-based reasoning programs with and without mental set support to examine their effects on conceptual change in tasks of varying difficulty. Additionally, we used an eye tracker to simultaneously capture students' attention and its sequential patterns during the conceptual change process, aiming to uncover the cognitive processes involved. Students were randomly assigned to either a computer-based reasoning program with mental set support or one without. We found that students in the reasoning group with mental set outperformed those without it in achieving scientific conceptual change. Furthermore, both groups demonstrated higher conceptual change success, along with longer fixation durations and rereading times, in low-difficulty tasks compared to high-difficulty tasks. Students who successfully achieved conceptual change exhibited more fixation points within areas of interest (AOIs) and significantly greater bidirectional connections between them. In contrast, students who failed to revise their alternative conceptions allocated fewer fixation points to AOIs and displayed significantly more unidirectional connections between them. Generalized estimating equation analysis identified the number of fixations as the most critical predictor of both conceptual change success and task difficulty. These findings provide new insights into the interactions between eye movement and underlying cognitive processes, highlighting how visual attention contributes to achieving successful conceptual change.