Nuclear data uncertainty is a major source of computational errors for key safety parameters in fast reactors. However, existing studies are largely confined to single or outdated evaluated nuclear data libraries, and most have focused on the steady-state eigenvalue k_eff, while systematic quantification of uncertainties in the kinetic parameter β_eff remains insufficient, and comparative analyses across different libraries are notably lacking. In this paper, a comprehensive uncertainty propagation study is performed for the China Experimental Fast Reactor (CEFR) based on the four latest-generation internationally mainstream nuclear data evaluation libraries: ENDF/B-VIII.1, JEFF-3.3, JENDL-5.0, and TENDL-2023. This work systematically reveals the distinct characteristics of fast reactor parameter uncertainties predicted by these state-of-the-art libraries and fills the research gap in this field. Methodologically, sensitivity coefficients for β_eff are derived using the k-ratio theory, thereby extending traditional first-order perturbation sensitivity/uncertainty (S/U) analysis framework from k_eff to β_eff. Concurrently, an energy-dependent uncertainty contribution factor method is proposed, which enables refined identification of the energy ranges, nuclides, and reaction reaction types that contribute most to the total uncertainty. All calculations are performed with our in-house code MCSU, which combines multi-group covariance matrices processed by NJOY2016 with adjoint-weighted sensitivities obtained from Monte Carlo transport calculations.
The results show that the four libraries yield relative uncertainties in k_eff ranging from 0.62% to 1.39%, with major contributions from the fission cross section, the capture cross section, and the prompt fission neutron yield of ²³⁵U. The relative uncertainties in β_eff range from 0.87% to 4.27%, predominantly governed by the delayed fission neutron yield of ²³⁵U, the prompt fission neutron yield of ²³⁵U and the delayed neutron yield of ²³⁸U. Notably, JEFF-3.3 severely underestimates the uncertainty in β_eff due to the absence of covariance data for the delayed fission neutron yield of ²³⁵U, highlighting the critical role of complete covariance information. Energy-dependent uncertainty contribution analysis further reveals that the intermediate- and high-energy region contributes the most to the uncertainties in both k_eff and β_eff, with fission cross section and delayed neutron yield of ²³⁵U dominating in this energy range. Significant differences in uncertainty predictions among libraries arise primarily from variations in covariance data, especially those of the ²³⁵U fission cross section and delayed neutron yield. This study systematically quantifies the k_eff and β_eff uncertainties for the CEFR based on multiple state-of-the-art nuclear data libraries, identifies priority directions for nuclear data improvement, and provides quantitative guidance for fast reactor safety assessment and future evaluated library development.