This study， based on data from the 2022 Hubei Province village household survey， evaluates the targeting bias in rural Dibao （minimum living allowance） by considering the current Dibao policy regulations and practices of income deduction as well as “single person assistance”. It examines the targeting bias of rural Dibao in terms of three types of Dibao eligibility criteria： pre-income for whole household assistance， pre-income deduction for household assistance and pre-income deduction for single person assistance. Using the double machine learning method， this study examines the improvement effect of rural governance innovation on the targeting bias of rural Dibao. The findings reveal substantial targeting bias， with previous research failing to account for income deductions and “single-person assistance，” leading to an underestimation of exclusion errors and an overestimation of inclusion errors.. Empirical tests demonstrate that rural governance innovations significantly and robustly improve targeting accuracy of rural Dibao. Therefore， it is recommended to enhance the refinement and promotion of rural governance innovations such as the points system， the list system， and digitization in rural governance， and to further improve the proactive identification and determination methods for assistance recipients.