Information problem solving (IPS) is an important twenty-first century skill, but it is lacking at all age levels. One type of information problem, those of an ill-structured nature that require multiple iterations of (re)defining problems and formulating emerging solutions, can be particularly challenging but have received less attention in the IPS literature. Further, the process of solving such problems often reveals, while simultaneously being impacted by, problem solvers' epistemic beliefs. Using a self-regulated problem-solving model as an analytic framework and taking advantage of multiple data sources, this study examined college students' self-regulatory patterns in performing an ill-structured IPS task, and compared the patterns displayed by two groups of students with more and less adaptive epistemic beliefs. Sequential analysis of behavioral data revealed different patterns between the two groups. Think-aloud data, interviews, and students' IPS products showed three key differences between the two groups: difference in the roles of IPS task instructions, difference in the numbers and triggers of queries, and qualitative difference in iterations between page viewing and writing. The findings yielded important insights into the self-regulatory processes of IPS and the role of epistemic beliefs at different problem-solving stages. Implications are drawn for educators and learning designers for developing IPS in higher education.