<![CDATA[Geometry problem-solving is an essential competency for Primary School Teacher Education students, yet many struggle with three-dimensional visualization. Furthermore, students' persistence in overcoming challenges known as adversity quotient (AQ) significantly influences this process. This study aims to develop augmented reality (AR)-based geometry problems with varying complexity tailored to students' AQ. Using the ADDIE research and development model with 29 student subjects at Universitas Muhammadiyah Malang, data were collected through observation, interviews, questionnaires, and documentation, then analyzed using descriptive quantitative and qualitative methods. Results show the developed AR-based problems are highly valid, scoring 74% from material experts (good category) and 96% from media experts (very good category), and yielding an 85.89% student response score regarding usability (Good Category). Crucially, empirical data regarding problem-solving skills demonstrated that AR integration successfully scaffolded early-stage spatial representation, enabling 82.76% of the subjects to overcome visualization barriers and comprehend three-dimensional contexts (Polya Stage 1). However, complete problem-solving success across all four Polya stages was achieved by only 24.14% of the cohort (7 students). The remaining 75.86% experienced structural drop-offs tied to specific AQ obstacles: Quitter students abandoned tasks at the initial stage due to persistent spatial barriers, whereas Camper students successfully devised plans but failed during the execution and verification stages due to arithmetic errors and a lack of looking-back habits. The varied problem complexity serves as an effective differentiated learning strategy, offering targeted visual scaffolding for lower-resilience students while demanding complex reasoning from higher-resilience ones. Therefore, this study contributes significantly to adaptive, technology-based, and inclusive geometry learning.]]>
