Assessing coupling coordination (CCD) between socioeconomic development (SD) and eco-environmental quality (EQ) is vital for sustainable development blueprints for mining towns. However, previous studies often overlooked the policy context and the asymmetry of SD and EQ impacts on CCD, in the modeling process, leading to a lack of objectivity and generalizability in the current methods. Therefore, this study devised a novel framework for assessing the CCD in mining towns by leveraging Convolutional Autoencoders and the Cusp Catastrophe model. The effectiveness of the framework was verified using Panxi mining town in China as a case study. Results demonstrate: (1) 65% primary coupling coordination indicates lagging SD in the Panxi mining towns in 2020; (2) there is an evident exponential growth trend in CCD of the Panxi mining towns grew exponentially (R2 = 0.94) from −0.351 in 2001 to 0.062 in 2020, and transforming from disorder to primary coupling coordination in 2015; (3) post-2013, effective local policies and measures boosted CCD by 35%, but currently they have not been continuously transformed into a driving force for CCD growth. Therefore, considering the practical challenges associated with different coupling coordination categories, this study provides recommendations for sustainable development at the township scale. The results provides insights for Panxi's sustainable development and aids in developing reliable predictive models for CCD for other mining towns.