Abstract: The overpressured reservoirs in the Huangliu Formation of the Dongfang area, Yinggehai Basin, are characterized by fine lithology, poor physical properties, complex storage space, and strong heterogeneity, which lead to poor correspondence between nuclear magnetic resonance (NMR) T₂ spectrum and capillary pressure curves, making the conversion between relaxation time and pore-throat radius challenging. To clarify the distribution relationship between NMR T₂ spectrum and pore-throat radius in such reservoirs and solve the difficulty in converting relaxation time to pore-throat radius, a segmented modeling method was adopted. Combining conventional physical properties, X-ray diffraction, NMR, and high-pressure mercury injection experiments, the conversion relationship between relaxation time and pore-throat radius was developed. The main controlling factors of surface relaxivity in different relaxation intervals were thoroughly analyzed. A four-factor regression analysis method was used to build a dynamic characterization model for surface relaxivity. The model achieved a relatively good calculation performance, with average relative errors of long and short relaxation components being 9.965% and 2.227%, respectively. The results indicated that the NMR T₂ spectrum and mercury injection-based pore-throat radius distribution curves of the reservoirs showed obvious segmented characteristics near 5.7 ms (corresponding to a pore-throat radius of about 0.1 μm). The surface relaxivity of short relaxation component (T₂ < 5.7 ms) was significantly lower than that of long relaxation component (T₂ > 5.7 ms), confirming notable differences in surface relaxation mechanisms across different pores sizes. Surface relaxivity was jointly influenced by reservoir physical properties, mineral composition, and pore structure. The surface relaxivity of short relaxation component was mainly affected by clay content, while the long relaxation component was mainly controlled by pore structure complexity. Based on this, a dynamic surface relaxivity characterization model was constructed through optimizing sensitive factors for different relaxation intervals, achieving quantitative prediction of surface relaxivity. Meanwhile, the correspondence accuracy between relaxation time and pore-throat radius was significantly improved through the segmented conversion method. This approach provides a theoretical basis and methodological support for the accurate conversion of pore-throat radius based on NMR T₂ spectrum. The findings are significant for precisely characterizing the pore structure of overpressured reservoirs and improving the accuracy of reservoir evaluation.