This study investigates the effects of a novel systematic approach to student-partner assignment for collaborative learning sessions in university Computer Science courses. The method utilizes assessments, specifically quizzes, administered in pairs, or dyads. Students complete the first quiz in a dyad independently and the second collaboratively with a partner. A student’s results from the first quiz determine their partner for the second, ensuring some degree of variability in the understanding of course concepts between the two students. The method is tested in multiple sections of two distinct university courses to explore whether student learning is influenced by the knowledge level of their partner, and whether the cumulative effect of partner assignments impacts a student’s overall performance in the course. One of the courses comprises first and second-year university students while the other is composed of third or fourth-year students. Statistical analyses, which utilize quiz-question embeddings generated by a large language model, proved inconclusive on the relationship between concept-specific knowledge and student learning. However, results indicate a positive correlation such that students who have partners with varying knowledge levels (relative to their own) performed better in the course overall. This relationship was statistically significant in courses for first and second-year students suggesting that deliberate partner assignment for collaborative learning may benefit students at the beginning of their university journey more than those further on in their academic journey.