Context: Food production in northern China faces persistent challenges from cold stress and water limitations, which threaten regional food security. To address this, plastic film mulching (PFM) is widely used in the region because of its dual effects of increasing temperatures and conserving soil moisture. Most assessments of PFM focus on crop yields, but research on the temporal yield stability across years remains insufficient. Methods: We compiled a dataset of 58 studies based on 87 comparisons to assess the temporal yield stability of maize and wheat in northern China. A machine learning model using 1089 yield observations was subsequently developed to generate maize and wheat yields across northern China during the historical (2000-2009) and future period (2031-2040), followed by an assessment of yield stability. Results and Conclusions: The results revealed that compared with the control, long-term PFM (>= 3 years) increased crop yields by 27% and resulted in significantly greater temporal yield stability (30%). With increasing water input, the maize relative yield stability initially increased but then plateaued in both the control and PFM systems. In addition, climate was a key factor affecting the relative yield stability ratio of PFM in maize, whereas it was primarily driven by N rates in wheat. PFM increased crop yields by 16.0-21.3% and the relative yield stability by 11.1-12.8% in northern China during 2000-2009. PFM achieved the greatest improvements in terms of both yields and relative yield stability in northwestern China, whereas its improvements were relatively limited in northeastern China. Furthermore, the yield and yield stability benefits of PFM were projected to decline under future climate conditions. In conclusion, PFM shows considerable potential for increasing yield and yield stability across northern China's cropland; however, its actual effectiveness is strongly influenced by local edaphic and climatic conditions. Sihnificance: This study provides the first rigorous assessment of how PFM influences crop yield stability using field experiments conducted over three years. By integrating long-term observations with a random forest algorithm, we identify nonlinear relationships between yield stability and key environmental drivers. Structural equation model further revealed distinct climate-soil-N fertilization pathways underlying the relative yield stability ratio under PFM in maize and wheat. Furthermore, we developed a machine learning model to evaluate its impact on yield stability across northern China, revealing regional heterogeneity and indicating that future climate change may reduce both yield gains and stabilizing effects of PFM relative to historical conditions.