Background and Objectives: Negative cognitive biases have been linked to anxiety and mood problems. Accumulated data from laboratory studies show that positive and negative interpretation styles with accompanying changes in mood can be induced through cognitive bias modification (CBM) paradigms. Despite the therapeutic potential of positive training effects, few studies have explored training paradigms administered via smartphones. The current study aimed to compare the effectiveness of three different types of training programmes (cognitive bias modification-attention, CBM-A; cognitive bias modification-interpretation, CBM-I; attention and interpretation modification, AIM) administered via smart-phones by using a control condition (CC). Methods:Seventy-six undergraduate participants with high social anxiety (Liebowitz Social Anxiety Scale, LSAS ≥ 30) were randomly assigned to four groups: CBM-A (n = 20), CBM-I (n = 20), AIM (n = 16), and CC (n = 20). Results: The results showed that the effects of CBM training, CBM-I training, or AIM training vs. CC for attention yielded no significant differences in dot-probe attention bias scores. The CBM-I group showed significantly less threat interpretation and more benign interpretation than the CC group on interpretation bias scores. Conclusions: The present results supported the feasibility of delivering CBM-I via smartphones, but the effectiveness of CBM-A and AIM training via smartphones was limited.