Accurately simulating global surface ozone has long been one of the principal components of chemistry-climate modelling, but divergences in simulation outcomes have been reported as a result of the mechanistic complexity of tropospheric ozone budget. Settling the cross-model discrepancies to achieve higher accuracy thus is a task of priority. Building on the Coupled Model Intercomparison Project Phase 6 (CMIP6), we have transplanted a conventional ensemble learning approach, and also constructed an innovative 2-stage enhanced space-time Bayesian neural network to fuse an ensemble of 57 simulations together with a prescribed ozone dataset, both of which have realised outstanding performances (R-square > 0.95, RMSE < 2.12 ppbV). The conventional ensemble learning approach is computationally cheaper and results in higher overall performance, but at the expense of oceanic ozone being overestimated and the learning process being uninterpretable. The Bayesian approach performs better in spatial generalisation and enables perceivable interpretability, but requires heavier computational burdens. Both of these multi-stage learning-based approaches provide frameworks for improving the fidelity of composition-climate model outputs for use in future impact studies.