By analyzing the current situation of local R&D talent projects in China, the general practices of local R&D talent project review were summarized, and the problem of differences in the assessment subject between groups that current peer review activities face was revealed. In order to equilibrating group error of assessment subject, a combined solution containing pre-review prevention and post-review correction was proposed. Before the review process, methods such as setting evaluation criterion and expert weights could be applied. After the review process, rank statistics method could be used to convert original scores into normally distributed rank scores for equalization, and then linearly transform rank scores back to the original interval distribution, which would effectively balance the differences in assessment subjects within and between groups. Random simulation and empirical practices show that this method can balance the differences of evaluation subjects between groups, which will help improving the objectivity of the score processing and providing certain reference to future local talent project review.