<p>Academic paper recommendation systems are vital in streamlining research by helping scholars identify relevant literature efficiently. However, traditional approaches often fail to capture profound contextual relevance, especially in multidisciplinary domains. This study proposes a novel Retrieval-Augmented Generation (RAG) model that synergistically combines retrieval and generative mechanisms to address these limitations. The RAG architecture, including data preparation, model training, and integration of neural retrieval and generation components, is described in detail. Experimental results demonstrate that the RAG model significantly outperforms standard content-based filtering, achieving precision of 0.78, recall of 0.72, F1-score of 0.75, and MRR of 0.83. These results validate the model’s ability to generate contextually relevant and accurate academic recommendations. This paper identifies future directions to further enhance the model’s applicability across disciplines.</p>