According to the grading standards for middle school examinations, thought content and structural coherence are critical indicators of essay quality. Evaluating an essay’s coherence requires not only analyzing the semantic connections between sentences but also understanding the logical relationships between sentences and the overall structural layout of the essay. This understanding is essential for grasping the logical framework of long texts and generating complex documents. This paper presents our approach used in the NLPCC 2024 competition for evaluating essay sentence logic and generating feedback. For different tasks, we applied both traditional classification models and advanced large language models, leveraging pseudo-labels generated through semi-supervised learning and information provided by large models as auxiliary data, significantly enhancing overall model performance. Experimental results demonstrate that our method achieves state-of-the-art results.