Abstract: To address the issues of information overload, lack of content matching and insufficient interpretability of the educational exam’s models of high school English multiple-choice reading comprehension, the model integrating knowledge augmentation and contrastive learning (KCL-MCRC) was proposed. Firstly, the Large Language Model (LLM) was employed to generate summary and extract clues from article for knowledge augmentation to reduce the inputs overload. Subsequently, the embedded vectors of article, question and options were represented by Bidirectional Encoder Representations from Transformers (BERT). The information difference vector was obtained through contrastive learning strategy combining mean pooling and absolute difference, which were subsequently concatenated with inputted embedded vectors to jointly impact the answer. Finally, the performance of KCL-MCRC was evaluated on each question type annotated by LLM to analysis the raising reason. The series of experiments were conducted on RACE-H dataset, including model comparisons, an ablation study of KCL-MCRC model, the effectiveness comparisons across question types, and an ablation study of its knowledge enhancement and contrastive learning modules for each question type. The results indicate that KCL-MCRC achieves optimal overall accuracies of 65.75% on validation set and 64.12% on test set, surpassing the second-best model called Multi-Decision-Transformer (MDT) and Multi-stage Maximization Attention (MMA) by 1.02, 0.88 and 0.35, 0.92 percentage points, respectively. Both knowledge enhancement and contrastive learning are crucial for model’s performance improvement, with significant accuracy gains in three question types: logical reasoning, main idea summarization, and author’s attitude. This validates effectiveness of designing modules specifically for different question types.