Abstract: To address the problems of insufficient utilization of single-field information, strong ambiguity in gravity and magnetic anomaly interpretation, and difficulties in basement lithology identification under complex geological conditions, this study proposes a gradient-based adaptive clustering method for gravity and magnetic multi-attribute interpretation. The method integrates multiple potential-field attributes, including the first vertical derivative of Bouguer gravity anomalies, reduction-to-the-pole magnetic anomalies, and their horizontal gradient magnitudes. A unified feature space is constructed through logarithmic enhancement and standardization, and the ISODATA algorithm is then applied to achieve adaptive clustering of gravity and magnetic anomalies. A series of model experiments involving different physical property contrasts, source burial depths, spatial proximity of anomalous bodies, and noise interference demonstrate that the proposed method effectively enhances anomaly boundary features and stably distinguishes different types of geological bodies. Moreover, it maintains robust boundary recognition and anti-noise performance even under complex anomaly superposition and noise interference. Application of the proposed method to gravity and magnetic data from the Weihe Basin yields clustering results that show good correspondence with the distribution of tectonic units, basement lithology, and existing geological knowledge, demonstrating that the method effectively reduces the ambiguity of single-field interpretation, improves the accuracy of integrated gravity and magnetic interpretation in complex regions, and provides a new technical approach for basin structure investigation and basement lithology identification.