Abstract: To solve the problem of difficulty and low inheritance efficiency in extracting tacit knowledge from tobacco grading experts, this study employed eye tracking technology based on visual research to explore the visual cognitive rules during tobacco leaf grading, extract expert tacit knowledge, and develop and verify the intervention effect of training materials of eye movement modeling examples (EMMEs). Fifty-five tobacco leaf grading experts were recruited, and their eye tracking data during single-leaf and tobacco bundle grading were collected using an eye tracker. Statistical methods were employed to analyze the eye movement patterns and extract tacit knowledge including attention allocation and gaze paths to construct EMMEs training materials. Seventy trainees were enrolled to conduct an intervention experiment with a 2 (high level/low level) × 2 (tacit knowledge intervention/conventional knowledge intervention) experimental design to verify the effectiveness of the training materials. A stable eye movement pattern of the experts was summarized. For single-leaf grading, leaf veins were the core, with the gaze path of “leaf vein→leaf surface→leaf tip→leaf base”. For tobacco bundle grading, leaf veins remained the core, while the attention proportion on the leaf surface was relatively higher, with the core gaze path of “leaf surface → leaf vein → leaf base → leaf tip”. The EMMEs training materials significantly shortened the single-leaf grading time of low-level trainees and improved their recognition accuracy of leaf position and grade, while only reduced the grading time for tobacco bundle grading. No significant improvements were observed in the grading time and accuracy of high-level trainees. Two core types of tacit knowledge, attention allocation and gaze paths, were extracted from the expert eye tracking data, and the visual cognitive rules of singleleaf and tobacco bundle grading were clarified. The EMMEs training materials constructed based on eye movement patterns improved the tobacco leaf grading performance of low-level trainees.