Abstract: Against the backdrop of China's transition into a “new normal” of sustained low interest rates,its government bond market—now the world's second-largest—has witnessed unprecedented expansion (with outstanding bonds surging from RMB 23.27 trillion in 2021 to RMB 34.57 trillion by the end of 2024) alongside heightened price volatility. This volatility has exposed the failure of traditional pricing mechanisms,as conventional linear models and single-cycle analyses can no longer capture the market's evolving logic. Addressing this critical gap,this study systematically explores the structural transformation of government bond pricing drivers and their nonlinear effects across interest rate cycles,delivering both theoretical breakthroughs and practical insights. This study clearly distinguishes between “normal-interest-rate periods” and “low-interest-rate periods” using a globally accepted criterion:a sustained overnight Shanghai Interbank Offered Rate (SHIBOR) below 2%. Drawing on daily data of key financial indicators from March 2018 to April 2025,it splits the sample into two phases:the normal-interest-rate period (March 2018–September 2023) and the low-interest-rate period (October 2023-April 2025). This division effectively captures China's first-ever sustained low-interest cycle,which is driven by cumulative monetary easing measures. The dependent variable in the study is the CSI Long-Term Government Bond Index,a representative indicator of China's government bond market. Independent variables include SHIBOR (reflecting market liquidity and funding costs),the onshore RMB index (indicating cross-border capital flows),the gold index (representing safe-haven demand),the crude oil index (linked to inflation expectations),and the Shanghai Composite Index (reflecting market risk appetite). Methodologically,the study adopts an innovative hybrid approach:it uses linear regression (including cointegration tests and correlation analysis) to identify linear relationships between variables across different interest rate cycles,and an artificial neural network (ANN) — equipped with a hyperbolic tangent activation function and the Levenberg-Marquardt algorithm — to capture complex nonlinear relationships. This hybrid method overcomes the limitations of traditional linear analysis frameworks. One of the study's most notable findings is the cyclical dependence of government bond pricing drivers. In the normal-interest-rate period,interest rates lose their explanatory power for bond prices (showing no stable cointegration with bond prices),while gold,the exchange rates,and the stock index emerge as the dominant driving factors. In contrast,the low-interest-rate period triggers a paradigm shift:interest rates reemerge as a critical factor with a negative impact on bond prices,and the hierarchy of driving factors becomes gold>interest rates>exchange rates. Crude oil exerts a weak negative impact in both cycles,while the stock index maintains a consistently mildnegative effect on bond prices. Results from the ANN further reveal significant nonlinear effects that linear models fail to capture. For gold:initially,rising gold prices drive up bond prices,but this driving effect diminishes gradually; once gold prices exceed a certain threshold,gold's safe-haven attribute shifts toward a speculative one,thereby suppressing bond prices. For interest rates:short-term interest rate cuts lead to a slight decline in bond prices,attributed to “liquidity trap expectations”; however,deeper interest rate cuts trigger accelerated growth in bond prices as investors adjust their expectations. Overall,the ANN achieves far higher fitting accuracy than linear regression,demonstrating its superiority in capturing complex market dynamics. In terms of theoretical innovation,this study is the first to formally define China's low-interest cycle and systematically compare the drivers of government bond pricing across different interest rate cycles. This fills a gap in existing literature,which often overlooks structural changes in pricing mechanisms across cycles. Additionally,the study extends multi-factor asset pricing theory by integrating nonlinear dynamics,proving that the “importance of pricing factors” is not static but varies with changes in the interest rate environment.By combining linear regression (for baseline associations) with ANN (for nonlinearity),the study overcomes the limitations of single-method approaches,providing a more holistic toolkit for analyzing complex fixed-income markets. For investors,it identifies actionable thresholds (e.g.,gold's 2,800 level) and rate-cut dynamics to avoid nonlinear risks. For policymakers,it warns that low interest rates sharply increase bond market sensitivity to rates,narrowing conventional monetary policy space,and highlights the need to monitor gold and exchange rates as systemic risk transmitters. In conclusion,this study reshapes our understanding of Chinese government bond pricing,proving that the market's logic undergoes a paradigm shift in low-interest environments. Its findings are not only academically significant for global fixed-income research but also practically critical for safeguarding China's financial stability amid monetary easing.