The surface atmospheric temperature (SAT) has experienced a significant increase with increasing anthropogenic greenhouse gases (GHG) since the Industrial Revolution. The polar regions have long been expected to warm strongly, making it a robust feature called polar amplification (PA). In particular, the Arctic region warms 2-3 times faster than the global mean temperature in recent observation and model simulations, referred to as Arctic Amplification (AA), with decreasing Arctic sea-ice extent and melting of the Greenland ice sheet. However, there is a strong warming asymmetry between Arctic and Antarctic: the Antarctic has yet to be observed and is projected to be weaker surface warming than Arctic with slightly increasing sea ice extent in the Antarctic. It is not easy to untangle the source of asymmetric surface warming between two polar regions. The CMIP6 offers an opportunity to identify the major factor contributing to warming asymmetry at the poles and its inter-model spread in a new model ensemble. In this study, we use a climate feedback-response analysis method (CFRAM) to decompose the surface warming into individual contributions of external forcing and internal feedback processes. Using the approach, we reveal that the seasonal energy transfer mechanism (SETM) is dominant at both poles and it plays an important role in multi-model ensemble (MME) polar warming asymmetry and inter-model spread. In summer at the poles, the surface albedo feedback plays a leading role in polar warming: solar radiation reaches its maximum and causes sea ice to retreat and water is exposed to solar radiation. Then the ocean acts as a reservoir for heat gained from solar radiation. The enhanced open ocean in the Arctic increases the upward sensible and latent fluxes to warm the winter surface. However, the Southern Ocean will take away part of the heat through the equatorward Ekman transport and heat released by ocean will decrease. Hence winter warming in the Arctic is much greater than in the Antarctic. Meanwhile, the strength of SETM in climate models is an important factor in determining inter-model spread of warming asymmetry at poles. The understanding of SETM will improve our standing of the physical processes that drive polar amplification, warming asymmetry, inter-model spread and climate impacts.