<![CDATA[This study examines deep learning—understood here as a pedagogical approach oriented towards meaningful understanding rather than as an artificial-intelligence technique—as a catalyst for competency-based educational transformation within Indonesia's Merdeka Belajar curriculum. Although the curriculum has been implemented progressively since 2021, classroom practice often remains procedural and examination-oriented, and Indonesian students' PISA 2022 performance remains below the OECD average. Using a qualitative systematic literature review guided by PRISMA 2020, we retrieved peer-reviewed articles, policy documents, and official reports published between 2020 and 2025 from Scopus, ERIC, DOAJ, Google Scholar, and SINTA-indexed journals; screened them against explicit inclusion and exclusion criteria; and analysed them through directed content analysis. The synthesis identifies three interrelated findings: (1) educational transformation stalls when learner-centred policy is not translated into metacognitive and assessment practices at classroom level; (2) the Merdeka Belajar curriculum provides structural affordances for deep learning—essential content, differentiated instruction and project-based learning—but these remain under-exploited; and (3) six principles of deep learning are associated with cognitive, affective and socio-psychomotor competence development. The study proposes an integrative conceptual model linking deep-learning principles with the philosophical and structural dimensions of the Merdeka Belajar curriculum, positioning metacognitive reflection, interdisciplinary integration and authentic assessment as the mechanisms through which curricular flexibility becomes substantive transformation. The model offers a basis for teacher professional development and for future empirical testing.]]>
