<p>Automated essay trait scoring involves a fine-grained evaluation that distinguishes it from the current mainstream target of evaluating essays holistically. Much of the existing research has focused on modeling distinct traits in isolation, often neglecting the interconnections between these traits and resulting in inefficient resource utilization and time wastage. Most neural network-based scoring models are data-driven, transforming scoring into classification or regression tasks, lacking explicit knowledge about the specific trait under evaluation. In this paper, we propose a unified framework for concurrently assessing multiple traits, incorporating trait description representation and a bidirectional attention of trait and essay, thereby augmenting the capability of the model to comprehend diverse traits. Experimental results demonstrate that the model we propose outperforms the baselines in average quadratically weighted kappa (QWK) for all traits on the ASAP/ASAP++ dataset by 3.3%. Additionally, our model exhibits enhancements in terms of both runtime efficiency and reduced model parameters.</p>