STEM Problem solving necessitates a substantial amount of specialized domain knowledge. An important element of problem-solving within the domain includes how knowledge is structured and organized in memory to facilitate efficient retrieval of relevant information and future problem solving. In previous studies, however, problem solving and knowledge structure have been studied in relative isolation, resulting in viewing them as a separate two-way process. To address this gap, this study aimed to track how individuals developed their knowledge structures before, during, and after collaborative problem solving in the contexts of STEM (physics, astronomy, and biology), with particular attention to understanding the different mechanism between success versus less successful problem-solvers. For that, we employed a relatively new and promising network approach to representing learners' knowledge structures as network graphs for analysis and comparison. Results visually demonstrated that successful problem-solvers tend to share a "solution-focused" knowledge and establish their "group knowledge-oriented" knowledge structure, whereas the less successful problem-solvers tend to share "problem-focused" knowledge and then establish their "prior knowledge-oriented" knowledge structure. Implications and discussion for the findings are provided.