Grasslands are the most extensively distributed terrestrial ecosystems on Earth, providing a range of ecosystem services that are vital for sustaining human life and critical for sustainable development at the global scale. However, the relationship between the two most important attributes of grassland, plant diversity, and productivity, remains controversial even after many years of research. Here, we develop an analysis of covariance (ANCOVA) model based on decadal-scale experimental data from a degraded meadow steppe in northeastern Inner Mongolia, China to quantify the response of aboveground biomass (AGB) to plant species diversity under varying management regimes. We report that AGB responds negatively to the plant diversity in fallow grasslands and positively in grazing grasslands, transiting from negative to positive in mowing grasslands as mowing became more frequent. We show that the changing diversity-productivity relationships are driven by changes in species composition of the plant community, given the significant productivity gap between rare and non-rare species. This highlights the role of management in regulating the diversity-productivity relationships in grasslands. These results not only provide provocative insights into the relationships between plant diversity and productivity but also support more sustainable use and management of grassland resources.