Learning-related attention is one of the most important factors influencing learning. Although technologies have enabled the automatic detection of students' attention levels, previous studies mainly focused on colleges or high schools, lacking further validations in primary school students. More importantly, the detected attention might fail to be learning-related if students did not attend learning tasks in the first place, which is common in real-world learning. This phenomenon poses challenges to the practical application of automatic attention detection, especially at the primary school stage, which is crucial for students to develop learning attitudes/strategies. Inspired by the emerging inter-person perspective in neuroscience, we proposed an inter-brain attention coupling method to detect learning-related attention as an extension to the existing single-person-based method. To test this method, wearable electroencephalogram devices were used to monitor students' attention levels in a primary school classroom. We found that one's inter-brain attention coupling, defined as the degree to which an individual student's attention dynamics match the attention dynamics averaged across other classmates, was positively correlated with academic performance: higher performances are associated with higher coupling to the class-average attention dynamics. Moreover, the attention detection framework based on the inter-person perspective outperforms as an indicator of academic performance compared with the widely-used attention level within an individual. The results provide practical insights by extending the applications of detected attention levels from an inter-person perspective and demonstrating its feasibility in monitoring learning-related attention among primary school students.