Help-seeking is an important self-regulated learning strategy and skill for effective learning. Studies have found that some students have poor help-seeking behaviors and that this leads to poor learning performance. Some researchers have developed help-seeking regulation mechanisms to detect and regulate students' poor help-seeking behaviors. Studies have also found that students have different help-seeking tendencies. Thus, adaptive help-seeking regulation mechanisms for different help-seeking tendencies are required. This study applied a help-seeking questionnaire and a K-means clustering approach to identify three help-seeking tendencies in the context of a computer assisted learning system (CALS). Then, adaptive help-seeking detection and regulation mechanisms were developed for these three help-seeking tendencies. The regulation mechanisms also adopted historical student records of problem-solving and help-seeking data for each problem as parameters to account for the difficulty of each problem. Furthermore, an experiment was conducted with a control group and an experimental group. Students in the experimental group used a CALS with adaptive help-seeking regulation mechanisms, whereas students in the control group used a CALS without the regulation mechanisms and could seek help at will. The experimental results showed that students in the experimental group had better learning performance for difficult problems, better help-seeking behaviors (i.e., less executive help-seeking) for easy problems, and a higher ratio of solving problems by themselves without seeking help than students in the control group.