Abstract: In order to further explore the regional heterogeneity of carbon emission efficiency of the same type of geographical unit, this study employs the MinDS model considering unexpected output and Malmquist index approach to investigate the static and dynamic urban carbon emissions efficiency of the Yangtze River Basin and the Yellow River Basin during the period 2005-2017, and further examine the spatial agglomeration characteristics and evolution law of carbon emission efficiency in the two basins from the perspective of the inter- and intra-basin comparison. Also, the influencing factors of different city types are evaluatedby using the random effect panel data regression model. The results show that: (1) The average carbon emission efficiencies of the Yangtze River Basin and the Yellow River Basin were 0.785 and 0.747 respectively during the underlying period. In general, the carbon emission efficiencies of the two Basins show a U-shaped curve. Both of them were in the rising stage of the curve from 2012 to 2017, but still stayed at a relatively low-efficiency level. (2) The carbon emission efficiency of the Yangtze River Basin presents a spatial distribution pattern of ‘downstream > upstream > midstream’, which is low in the middle and high at both ends, and the carbon emission efficiency of the Yellow River Basin presents the spatial distribution pattern of ‘downstream > midstream > upstream’. The high value areas of carbon emission efficiency in the Yangtze River Basin show the trend of agglomeration, and the low value areas are more scattered. While the low value areas of the Yellow River Basin are centered on the Ningxia urban agglomeration along the Yellow River and spread along the main flow of the Yellow River, the high value areas are small and scattered. (3) The Malmquist index shows an upward trend. The technological progress index, which denotes technological innovation level, is the main factor contributing to the improvement of carbon emission efficiency in the two Basins. The technological efficiency index, which represents the combination of production factors and management level, has no significant effect. (4) Based on the difference in the role of technological efficiency and technological progress in the improvement of carbon emissions efficiency, the research objects could be divided into six types of cities. The level of economic development and industrial structure are common factors that affect the improvement of carbon emission efficiency in the two major river basins. The study shows that the changes of carbon emission efficiency in the two basins have both overall similarities and intraregional heterogeneity. We should consider not only the general impact of industrial structure on the improvement of carbon emission efficiency in the two river basins, but also the differentiated impact of factors such as the level of urbanization, to realize the ‘local conditions and classified measures’ in the policy design of carbon emissions reduction and efficiency improvement.